Multipath Simulation for SSB FM Stereo

[k6sti]

The simulation outputs a single channel. Normally I decode the modulated channel and see what multipath did to it. Occasionally I will flip the stereo matrix and pick up the other channel to check stereo separation. Whenever I've done that with multipath enabled, the amount of distortion in the unmodulated channel has impressed me. I've heard this effect on the air. I decided to put the two channels together and present them as stereo, the way you'd actually hear it from a receiver. When presented this way, the multipath character of the sound is unmistakable. You can hear what it sounds like by clicking on the updated sound sample.

The other thing I did was to generate a spectogram of the stereo audio signal. It's quite revealing, clearly showing how multipath distortion fills the voids between the voice formants and harmonics.

http://ham-radio.com/k6sti/ssb.htm

Brian

 

[k6sti]

Summing the noise (which really is distortion, but I can't bring myself to write S/D) in both channels rather than just the one modulated seemed like such a good idea that I redid the entire multipath simulation writeup using that approach. This better reflects what a listener actually hears when experiencing multipath. The SSB S/N advantage numbers changed, but not enough to matter, I think.

I also added several more stereo sound samples of simulated multipath, not all of which may survive past April 1.

http://ham-radio.com/k6sti/ssb.htm

Brian

 

[FFoti1]

We did an actual full system test of stereo separation on a Sony tuner to compare the simulated data Brian shared in his paper. We were able to find the same tuner he references and used it for the end-to-end test. The transmission system comprised of an audio oscillator to generate the single channel reference signal, an Omnia.11 (in bypass mode) for the DSB/SSB stereo generation, Nautel VS300 exciter, Sony tuner, and a Tektronix TM-504 measurement system.

Below are the results. The simulated separation numbers from Brian's paper are plotted under the columns labeled "simulated." The complete system test results are under the columns labeled "actual." There is a significant difference between the simulated results, and what a complete system test reveals. Broadcast engineering practice employs the complete system test, as compared to simulating separation based on the IF response of a tuner. Further, we tested a NAD 4225 tuner, about the same vintage as the Sony tuner, and the results are about the same as the "actual" results we found with the Sony.

Freq DSB-Wide (simulated) DSB-Wide (actual) DSB-Narrow (simulated) DSB-Narrow (actual)

1kHz -70.0 -40.5 -48.0 -43.8
2kHz -56.0 -41.5 -42.0 -45.0
3kHz -50.5 -42.0 -39.5 -45.8
4kHz -49.0 -42.2 -38.0 -46.3
5kHz -48.0 -42.3 -38.0 -46.6
6kHz -47.5 -42.3 -38.5 -46.5
7kHz -47.5 -42.3 -39.5 -46.5
8kHz -47.5 -42.2 -40.0 -46.1
9kHz -47.5 -42.2 -42.0 -45.4
10kHz -48.0 -42.2 -46.0 -45.4
11kHz -49.0 -42.1 -52.0 -45.2
12kHz -55.0 -41.9 -53.0 -43.7
13kHz -53.0 -41.8 -47.0 -44.0
14kHz -56.0 -41.6 -42.0 -44.5
15kHz -56.0 -41.6 -39.5 -44.5

Freq SSB-Wide (simulated) SSB-Wide (actual) SSB-Narrow (simulated) SSB-Narrow (actual)

1kHz -52.0 -33.0 -40.5 -40.4
2kHz -46.0 -33.4 -35.0 -42.5
3kHz -43.0 -33.6 -31.0 -40.7
4kHz -40.0 -34.0 -29.0 -39.3
5kHz -38.0 -34.1 -27.0 -38.4
6kHz -37.0 -34.2 -25.0 -37.9
7kHz -35.5 -34.4 -24.0 -37.5
8kHz -34.5 -34.4 -23.0 -37.1
9kHz -33.0 -34.4 -22.0 -36.6
10kHz -32.0 -34.6 -21.0 -36.6
11kHz -31.0 -34.6 -20.5 -36.4
12kHz -30.5 -34.7 -20.0 -36.1
13kHz -30.0 -34.8 -19.5 -36.1
14kHz -29.5 -34.8 -19.0 -36.2
15kHz -29.0 -34.8 -18.0 -36.2

Considering the large variance between the simulated and actual results, this calls into question the validity of his data in the paper. As stated in a prior reply, simulations offer an idea of what might occur, but there may be other real world variables that affect the actual results.

-Frank Foti

 

[k6sti]

1. You neglected to align the stereo separation of your tuner before you performed the test. Wide-IF separation should be at least 50 dB at 1 kHz when properly aligned. My calculations assume that this has been done.

2. My modified tuner has individual stereo separation adjustments for the wide and narrow IFs. The stock tuner has only one adjustment. My calculations assume that both adjustments have been made for best separation at 1 kHz. This isn't possible when there is just one control.

3. My tuner uses 110-kHz filters in narrow. The stock tuner uses much wider filters (I believe they are 250s). This discrepancy will make a large difference in narrow separation, both calculated and actual.

4. I made my wide composite measurement before the tuner's elaborate multipole/multizero postdetection filter. Most tuners do not do any postdetection filtering beyond that provided by an IF-ripple bypass cap. I wanted the composite measurements to apply to circuits more typical. The tuner you tested passes the composite through its postdetection filter. This will alter the amplitude and phase of the signal presented to the stereo decoder in ways my calculations can't account for since the measurements they are based on were made earlier in the signal path.

Brian

 

[FFoti1]

Brian,

I highly doubt the average broadcaster, or listener is going to modify their tuner, as you did yours. The point of this exercise was to observe separation results that were done on an actual transmission system, as compared to a simulated system. It's quite obvious there is a large difference between the two methods. The broadcast industry relies on actual end-to-end, complete system test measurements, not a one-off receiver only simulation. Your data is based on the latter.

As asked of you prior, your simulated test results need to be confirmed through a complete end-to-end system test. That being, source signal, stereo generator (DSB/SSB), exciter/xmtr, tuner under test, and measurement device. Then you can compare your simulated data with actual end-to-end data. I have observed other instances in communications where simulated data and actual data are significantly different.

Based on the description of your highly modified tuner, you are basing your data on a one-off receive only system, and this is not typical of what will be found in the marketplace.

-Frank Foti

 

[k6sti]

You compared apples and oranges, Frank. There is no reason the measurements you made should agree with my calculations. They do not apply to your tuner for the several reasons I stated.

If you'll take a look near the end of my writeup, I explain why DSB and SSB stereo separation must differ for any tuner with a nonflat composite response. The reason is really quite easy to grasp.

http://ham-radio.com/k6sti/ssb.htm

Brian

 

[FFoti1]

You miss the point. Calculations are nice, but they need to be representative of what happens in the real world. Thus far, your data does not support this. Now, we learn, you are gathering your information from a one-off modified tuner. Not a device commonly found in the marketplace, and one which could be verified by another source.

-Frank Foti

 

[k6sti]

Frank, I offered the stereo separation curves as examples to illustrate the principle explained in the previous paragraph. They are not meant to apply universally to all equipment. That's why I stated the measurement conditions in some detail, including the ceramic filter part numbers. It's not possible to calculate stereo separation for an arbitrary receiver with unknown alignment. Separating L from R is a sensitive subtractive process. There are way too many variables and unknowns that can only be resolved by making specific measurements. However, it is possible to predict in principle what will happen when an SSB signal meets a sloping composite response. The curves show one example.

The composite measurements I made are contained in the software utility referenced at the end of the section. It's easy to use the numbers and calculate the resulting stereo separation by hand. I recommend that you try this so you can see what happens in a nonabstract way.

I don't know whether the separation loss SSB entails is enough to inhibit its adoption. That's a complex question that I think can only be answered in part by extensive testing of many different receivers in the field. But you need to understand why SSB has an inherent stereo separation issue that DSB doesn't. It will certainly come up in any FCC rule-making process.

Brian

 

[fm-engineer]

This thread is growing old. The validity of SSB will be determined in actual transmission systems. People like me will turn on SSB; drive around in known trouble areas with different vehicle receivers, and make a final decision with the input of management. That is what most listeners do. Drive around to and from work while listening to the radio. No one cares about simulated lab tests with only one receiver. If you told me or the 17 stations that I work for we could improve signal performance in trouble areas, but there may be one or two Sony receivers that blend to mono with SSB, it would truly be a no brainer.

 

[FFoti1]

@fm-engineer, very well said.

This point about stereo separation is a bit overblown. To the ear, differences in perceivable stereo separation of 30dB, 40dB, or 50dB is hard to discern. Consider, if you will, most of us were willing to accept stereo separation of 25dB to 30dB using a phono cartridge. On a rare occasion, you could achieve 35dB separation. SSB separation is greater than most of those phono cartridge numbers.

If I had to choose between reduced multipath using SSB, but possibly giving up a bit of separation which most people will not hear the difference, or maintaining the status quo and accepting the existing amounts of associated annoying multipath, it's an easy decision...SSB wins. Less multipath equates to the ability of retaining listeners, which is the whole point of this exercise. The Boston demo on YouTube is case in point. In that instance DSB caused a tune out, whereas SSB enabled the signal to be listenable.

-Frank Foti

 

[WNTIRadio]

Half the time people's car radios are blended to mono.

If you have a station at your disposal and management willing to try this for a couple of days, leave the pilot on but with mono audio. See if anyone calls and complains about it. Chances are, you may get one or two. Chances are greater that nobody will notice.

I have two translators I take care of that run an AM, both are in mono. Nobody has even noticed.

Sure, it's great to hear a mod-monitor with 65dB of separation, but who really has that at their home besides some of us engineers and hams?

8VSB did great in the lab and sucks in the real world. It's passable on UHF, but try receiving any one of the VHF DTV stations more than 20 miles, and sometimes closer, to the transmitter. That doesn't mean that lab results aren't valid, but it does mean they are one small subset of numbers on something with VERY large variables. Multipath while stationary can fluctuate quite a bit, while moving it is a constantly changing source of interference. If there is a way to mitigate some of that perceived interference (remember, 99% of the public has no freakin clue what multipath is, they just know "my station is staticy") at the expense of a few dB of separation, then I'm on board to do it.

Now also imagine tuners that have filters tuned to SSB and how selective they could be...

This FCC is more concerned with broadband and selling off TV licenses than FM radio. I'm sure if engineering can prove a reduction in multipath but a few dB are shaved off the total separation, it will pass. If a station wants to use it, they can. If not, then they can continue in DSB.

 

[FFoti1]

Frank, I offered the stereo separation curves as examples to illustrate the principle explained in the previous paragraph. They are not meant to apply universally to all equipment. That's why I stated the measurement conditions in some detail, including the ceramic filter part numbers. It's not possible to calculate stereo separation for an arbitrary receiver with unknown alignment. Separating L from R is a sensitive subtractive process. There are way too many variables and unknowns that can only be resolved by making specific measurements. However, it is possible to predict in principle what will happen when an SSB signal meets a sloping composite response. The curves show one example.

The composite measurements I made are contained in the software utility referenced at the end of the section. It's easy to use the numbers and calculate the resulting stereo separation by hand. I recommend that you try this so you can see what happens in a nonabstract way.

I don't know whether the separation loss SSB entails is enough to inhibit its adoption. That's a complex question that I think can only be answered in part by extensive testing of many different receivers in the field. But you need to understand why SSB has an inherent stereo separation issue that DSB doesn't. It will certainly come up in any FCC rule-making process.

Brian

The problem with your example is you have not confirmed if your test method replicates the actual results of an end-to-end test. You are using a highly modified one-off tuner, which has not been tested end-to-end. Until you can provide actual test results, which support your data, your one-off setup appears to be an unknown entity. Some might consider it hobbyist in nature. Good engineering practice requires criteria which can be replicated by others.

Regarding further testing, I have been a strong proponent of further testing. At issue here, which impacts future testing, is the use of known end-to-end test methods, as compared to yours, which have not been confirmed to reproduce the same results of a complete system test.

-Frank Foti

 

[k6sti]

The problem with your example is you have not confirmed if your test method replicates the actual results of an end-to-end test. You are using a highly modified one-off tuner, which has not been tested end-to-end. Until you can provide actual test results, which support your data, your one-off setup appears to be an unknown entity. Some might consider it hobbyist in nature. Good engineering practice requires criteria which can be replicated by others.

Regarding further testing, I have been a strong proponent of further testing. At issue here, which impacts future testing, is the use of known end-to-end test methods, as compared to yours, which have not been confirmed to reproduce the same results of a complete system test.

-Frank Foti

Frank, I wasn't "testing" my tuner. I don't have an SSB generator. Instead, I measured the amplitude and phase of the tuner's composite response. Knowing that, it is possible to calculate how any tuner will respond to a DSB or SSB signal. It doesn't matter what the particular composite response is, whether it is standard or special, whether the IF filters are wide or narrow. Given the composite magnitude and phase of any receiver, it possible to determine what the stereo separation will be at any frequency. If this is news, perhaps a technical person at Omnia can go through the details with you and explain how this is possible. The value in doing this kind of calculation is that it will become obvious why a DSB signal can accomodate a sloping composite response while an SSB signal can't. This kind of response is characteristic of analog IF systems.

SSB stereo signals are not fully compatible with today's stereo receivers, and degradation of stereo separation is only one reason. Whether an SSB signal might provide enough benefit to overcome its inherent disadvantages remains to be demonstrated.

Brian

 

[FFoti1]

Brian,

I understand what you are saying, and the method you refer to. You need to prove your validity by comparing the results of your method to a complete end-to-end test. Until you can offer that information, your method does not substantiate known broadcast practices, nor does it take into account the rest of the system. As stated prior, there are many examples of test setups, like yours, which offer one set of data, but an actual and complete real world test setup offers another. We need test results which represent the actual system performance, not something that is ad-hoc.

Even the receiver manufacturers use a complete end-to-end test to verify performance. You are basing your criteria on a single-ended (receive only), ad-hoc setup. You offer data that is not complete or representative of what would be found in the marketplace, as you are modeling a modified one-off tuner.

-Frank Foti

 

[FFoti1]

1. You neglected to align the stereo separation of your tuner before you performed the test. Wide-IF separation should be at least 50 dB at 1 kHz when properly aligned. My calculations assume that this has been done.

Brian

I reset the alignment of the tuner to maximize separation (>50dB) at 1kHz, and ran the separation response plots again. This was done using the complete end-to-end system test. In wide-IF and narrow-IF, stereo separation parallels the response plots I shared earlier. This is true for both DSB and SSB transmit modes. Again, this is in conflict with your calculated response plots, and further supports the complete end-to-end test, as compared to a calculation method, which is based upon the receive-only IF response of the tuner. This now calls into question the credibility of your data, as it does not support a standardized accepted practice.

-Frank Foti

 

[k6sti]

Frank, there was nothing ad-hoc about my test, it was not receive-only, it did not violate ordinary engineering practices, and I was not "modeling" anything. I transmitted a signal to the tuner using a laboratory signal generator and measured its composite response. From this the stereo separation can be deduced. I took special measures to exclude the tuner's complex postdetection filter when measuring the wide-IF response since most receivers do not use one. Although it was very tedious to do, I also went to the trouble of measuring composite phase, which for this receiver turned out to have little effect.

The measurements I made were for just one tuner. I present them as an example, not as a comprehensive final word on the matter. I believe you're critical of my test not because it was invalid, or because it was not comprehensive, or because it was done in a lab. I think you simply don't like the results. I believe the same is true for my multipath simulation. Had either shown a strong advantage for SSB, I have no doubt you'd be praising them to the moon.

Brian

 

[FFoti1]

Brian,

Your methodology does not reproduce the same results as a known, standardized, and accepted method. If it did, then I'm sure others would employ it, including me. If the results of the end-to-end method agreed with your findings, then this entire discussion would be moot.

You are presenting data, even in the manner you describe, which does not equate to what an actual complete transmission system creates.

-Frank Foti

 

[satech]

This point about stereo separation is a bit overblown. To the ear, differences in perceivable stereo separation of 30dB, 40dB, or 50dB is hard to discern. Consider, if you will, most of us were willing to accept stereo separation of 25dB to 30dB using a phono cartridge. On a rare occasion, you could achieve 35dB separation. SSB separation is greater than most of those phono cartridge numbers.

But is the reduction in stereo separation caused by SSB only limiting the maximum possible separation on an ideal tuner, or is it across the board? Because if it is across the board, then an average consumer-grade tuner getting reduced from 30 dB to 15 dB of separation is going to be much more noticeable and objectionable than a high-end tuner getting reduced from 50 dB to 30 dB of separation.

 

[FFoti1]

But is the reduction in stereo separation caused by SSB only limiting the maximum possible separation on an ideal tuner, or is it across the board? Because if it is across the board, then an average consumer-grade tuner getting reduced from 30 dB to 15 dB of separation is going to be much more noticeable and objectionable than a high-end tuner getting reduced from 50 dB to 30 dB of separation.

We've tested a variety of tuners and receivers, from varying price points. Stereo separation in SSB mode does not change in an "across the board" manner. We have some test results, from a few receivers that have no change at all between DSB and SSB.

-Frank Foti

 

[RealityCheckr]

Frank, there was nothing ad-hoc about my test, it was not receive-only, it did not violate ordinary engineering practices, and I was not "modeling" anything. I transmitted a signal to the tuner using a laboratory signal generator and measured its composite response. From this the stereo separation can be deduced. I took special measures to exclude the tuner's complex postdetection filter when measuring the wide-IF response since most receivers do not use one. Although it was very tedious to do, I also went to the trouble of measuring composite phase, which for this receiver turned out to have little effect.
<snip>
Brian

But what is the point? There is no point other than to prove you are good at writing software. But even that is suspect unless your model is shown to represent real results, something you refuse to do.

jeepers