Does the ARRL Bandwidth Plan Go Far Enough?

Ok, if anyone is still reading this thread here is a summary of what I put together for a test. I can't complete the testing until I get my old tempo 2020 to working on SSB. It seems to tune up fine but the audio is very distorted which doesn't work for a test.

First, let me say that a full, repeatable test of the compatibility of analog and digital is beyond my basic laboratory.

One major problem is comparison of actual transmitted signals. I don't have a high power signal combiner that will let me combine two signals of significant enough power to allow measurement using my monitor scope or my oscilliscope. Nor do I have a spectrum analyzer that will let me look at received signal levels to do an accurate comparison.

I therefore am forced to use the S-meter on my test receiver and that is a difficult tool to use on a repeatable basis. According to my signal generator and a stepped attenuator, the S-meter on my test receiver has a signal change of about 3db per division in the middle of the range between S3 and S7. Beyond S7 the scale is too compressed to be of much use.

My test receiver also has only a 1.8khz SSB filter available which also makes a complete comparison difficult since the entire bandwidth of a typical SSB signal can't be used. I have the crystals available to build wider filters, I just haven't had the time to do so. The receiver also has a very fast AGC attack and release time that is not adjustable. How these results would change on a receiver with a slow release AGC I cannot estimate.

Another caveat is that the words I used were recorded onto a small tape recorder using micro-cassettes and played back through the headphone jack. I have not measured the bandwidth of the recorder but I suspect it is limited.

A full analysis of the speech intelligibility would require at least 5 different speakers and 5 differnet listeners and a statistical analyis of the results. Since I don't have that capability here, at least not without getting a lot of friends to come over, I decided to use a more subjective measurement of the type used for telephone circuits called the Mean Opinion Score (i.e. bad =1, 2=poor, 3=fair, 4=good, 5=excellent). Once again, a full measurement would require several listeners to perform the test and the results would be averaged. I am only going to give my score. I used words from a standard ryhming test to record word pairs  such words as went/rent, pat/pad, din/dim, led/fed, sum/sun, wig/rig, mass/math, etc.


Test 1: I setup the multipsk program to send mfsk16 and adjusted the interface level and microphone level to about the S5 level while mfsk16 was idling and sending no data. I then sent a string consisting of CQ eight times and my callsign 3 times. The S-meter showed a very tiny oscillilation during idle and a swing of slightly below S5 to about S6.5 during the send data. This is a change of 4-5db. I then adjusted the SSB transmitter using the microphone gain to hit S7 when sending a 750hz tone into the microphone that had been recorded on the tape recorder. When playing the recorded voice into the transmitter, peak levels of S7 were seen.

This is as far as I could get. Even with the digital signal off I could not understand what was coming through the receiver well enough to distinguish words in the rhyming test. I'll either have to get my 2020 fixed or borrow another transceiver to use in the testing.

It is obvious that baseband testing is not going to give adequate results. When an mt63 signal shows barely 5watts output on the transceiver yet the receiver shows a consistent S7 and an mfsk16 signal with 15 watts out has a 10over S9 signal while a SSB signal of equivalent peak readings has the Smeter jumping from S7 to zero during speech, it is obvious that there is enough signal power difference in the IF amp stage of the receiver that only RF level receiving tests are going to be adequate to truly judge impacts.

I know enough now to know that this testing *should* be done somewhere, somehow, somewhen by someone - preferably with a fully equipped lab. There should be at least three separate station setups capable of SSB and various data modes. Full testing of all combinations should be done. Without this kind of knowledge it is just impossible to make rational decisions concerning the mixing of analog and digital signals. The only basis we have right now is the empirical observations done in Region 1, and while I give this significant weight, others apparently do not.

It *should* have been done, or at least sponsored, by the ARRL before their proposal of mixing analog and digital was made. Either they were too lazy to do the testing, they were not technically astute enough to even realize the testing needed to be done, or they didn't really want to know the results. Which ever it was doesn't matter to me, they all show that the ARRL can't be trusted to provide their constituents enough information to make rational, informed judgements. That also means the ARRL itself is making major decisions concerning the future of the amateur service based on the reading of tea leaves instead of measured, repeatable test results.

I'll leave it to the readers to make a judgement on how poorly this reflects on the ARRL.

tim ab0wr

 

Thanks for the info Tim. I agree tests should have been done and published. We need to objectively know what the boundries are for cohabitation of modes. Your attempts are appreciated.

Perhaps the people who SHOULD be doing this test will pick up the ball.

 

I think you misunderstand. Because of path loss on HF, several SSB conversations can occupy the same frequency at the same time as long as they are separated geographically.

How many conversations can be involved in this "propagation stacking" is dependent upon several factors.

It is not intuitively obvious that digital modes can offer this level of spectrum efficiciency. A human can distinguish between several SSB conversations operating at different dynamic levels. A digital mode usually is not set up to do so. If two digital signals exist on the frequency most digital programs will copy the strong one when it is on and the weak one when the strong one is off. You therefore wind up with intermixing of two digital streams which is "interference". Whether the receiving operator can distinguish between the transmissions is problematic. The answer is probably sometimes yes, sometimes no.

I know from empirical observation over many years that on 75m in the evenings it is entirely possible for conversations in the upper east US, the Southeast US, and the Central US to be in progress on 3920khz all at the same time and for each conversation to progress satisfactorily.

How well various types of signals can mix will be a determining factor for the spectrum efficiency of each type of signal.

While this can be done on a subjective basis using empirical observation it would be very nice to have repeatable measurements that can be used by anyone.

For instance, based on empirical observation I know that psk31 cannot exist with a SSB signal in the 350-800hz area of the SSB signal. There is just too much power in the SSB signal and you get big gaps in the psk31 received data stream while the person is speaking. On the other hand, psk31 signals in the 800-2500hz area can communicate very well but are very aggravating to the SSB operators and can impact the intelligibility of the SSB signal if they are in the 1500hz area or the 2200hz area because that area of the SSB signal carries a significant part of the intelligibility.

We *need* some absolute measurements of this from a fully equipped lab in order to establish operating parameters. Those parameters may turn out to be so restrictive that it will mean the signals should not intermix in the same band segment.

That is certainly the recommendation of the Region 1 HF Managers and the Region 1 overall Plenary. They recommend against mixing analog and digimodes. But again, that is based on empirical observation and not on actual measurements.

So the thrust of this testing is not for placing responsibility for interference but to provide reasons for making judgments on bandplanning. These judgments should be based on spectrum efficiency. If SSB provides three times the spectrum efficiency of most digital modes, then mixing analog and digital becomes a questionable decision. If SSB provides even twice the spectrum efficiency, then an informed debate needs to be done. If SSB provides no spectrum efficiency over digital, then who cares?

I know what my judgement would be based on my empirical observations but it is not a guarantee that others will make the same judgement.

tim ab0wr