High Fidelity Audio

[k4kyv]

The subject of hi-fi SSB was brought up about a week ago on QRZ.com and did it push some buttons! The thread is now 17 pages long.
Read QRZ.com  thread

NU9N's website has an excellent discussion on the bandwidth issue regarding hi-fi SSB.  Many of the issues he brings up apply equally well to AM:

Page 1
Page 2


NU9N cites an interesting White paper on "The Effects of Bandwidth vs. Speech Intelligibility".  Click on the link below to view (PDF format):

http://www.nu9n.com/images/Sound.pdf

[Ian VK3KRI]

Unfortunately the defined bandwidth of the telephone ~ 300-3300Hz seems to have a magic appeal as 'the communication standard'. As the last article you linked to said , that has more to do with economics than anything else. All the data I've seen indicates the knee in the articulation index rolls off at somewhere about 7khz for male talkers. None of this is new, there's even a table in the Radiotron Designers handbook (4th ed). 

Just as important is distortion. Why some people think that turning up a compresor and producing what sureley must be just about bandwidth limited squere waves is beyond me. Yes its louder, but its LESS intelligable.  The signal that stands out on the band is the one sounds clear and balanced not the one that sounds like someones decided to try out a '70s transistorised fuzz box as a mic preamp.                         Ian VK3KRI

[Tom WA3KLR]

One thing I've noticed is that I never hear a ham complain that he has trouble communicating over the phones.

WE/Bell did much testing early on (1920's) on intelligibility: sound level, noise, S/N, frequency response, distortion, echo tolerance - delay time and level, etc.  As stated above, some of this data must be in the Radiotron Designers handbook (4th ed) and other engineering handbooks.

The 300 - 3300 Hz range was then picked with good statistical data and of course some amount of compromise for costs, but appears to do the job.  SSB voice communications uses even less bandwidth than the phone system.

[WA1GFZ]

Tom,
My new SDR toy with spectrum display shows that 300 to 3000 does the job but the guys who go  out 8 or 10 KHz sound much better and it is easier on the ears.
The only multiplexing going on is between the soldering iron and rx.

[KB2WIG]

One may also consider why women are prefered as public safety dispatchers.... among several things, its their voice characteristics ... klc

[WA1GFZ]

men pay attention to a woman's voice beause they wonder if they are as pretty as they sound.

[Jim, W5JO]

I have more trouble with women's voices for two reasons.  Their voices tend to vary acoss the spectrum much more tending more toward high frequencies and their propensity to become exicted which accentuates the first complaint.

Just frankly I dislike femaie newscasters.  Maybe they should be hosting cooking shows.  Bell telephone performed many tests and determined tha optimum frequency acorss the hearing range of most people was 1001 cycles.  Below about 300 cycles the found that it did not add to the intelligibility and used more power than the good it produced.  Mostly above about 4500 many people did not notice the missing frequencies.  So they cut it off where people began to complain about poor sound.  Today their tests would be called a focus group.

Radio added those upper and lower frequencies to accomodate music.  With much of the music broadcast today, I wish they would limit their frequency response to 300-3 Kc.  I would not miss a thing.

[k4kyv]

Bell telephone performed many tests and determined tha optimum frequency acorss the hearing range of most people was 1001 cycles.  Below about 300 cycles the found that it did not add to the intelligibility and used more power than the good it produced.  Mostly above about 4500 many people did not notice the missing frequencies...Radio added those upper and lower frequencies to accomodate music. 

The frequencies below 300 add natural quality to the voice, and make extended periods of listening less fatiguing.  They don't have to use an inordinate amount of power.

Let the frequency response run flat down below 100~  and begin boosting the highs somewhere about 800~.  Let the response curve rise steadily up to 9-10 dB at about 2000~, then flatten out and run flat to somewhere between 3500~  and 4500~.  That will produce good articulation response and thus good intelligibility under less than ideal conditions, but the signal on the other end will not sound bassy or tinny.

There is already some presence rise is built in to the D-104, but it needs a little help on the low end.  This can be accomplished in the speech amp, but I simply mix in a dynamic mic I happened to have on hand, that has good bass response but poor high frequency response.

To avoid undesirable phase shift distortion, the actual response capability of the speech amplifier-modulator should run at least one octave above and one octave below the intended frequency response of 100-4500~, that is have a capability of running  flat as measured with signal generator and scope over the range of at least  50-9000~.  That doesn't mean you have to actually transmit 50-9000~ of audio, but that the 100-4500~ audio for speech will sound cleaner, smoother and more natural, and the asymmetry will be preserved, if the mod iron and associated interstage transformers and coupling components have the capability of at least 50-9000 ~,  plus or minus a dB or so.  Most broadcast iron is good for 50-10K flat response or better, but often above 10K you begin to see serious phase shift and the response curve may become ragged.  I find this true with the UTC LS series transformers, even though the manufacturer's specs claim flat response out to 15K or 20K or more with some transformers.  But with voice, there is no real reason to transmit anything above about 5 kHz, since very few ham  receivers will reproduce that high anyway.

Sharply cutting off the audio below 300~ makes the voice sound thin and 1950's telephone quality.  I say 50's quality because some of today's phones exceed the traditional 300-3000~ response and actually sound pretty good. But to get an idea of what "telephone quality" really sounds like, listen to some of the AM talk shows and compare the quality of the announcer's voice to that of the caller.

I'd rather sound like the announcer than like the caller.

[wavebourn]

With much of the music broadcast today, I wish they would limit their frequency response to 300-3 Kc.  I would not miss a thing.

Narrow band at 63 Hz for bass drum is fairly enough. 

[Bacon, WA3WDR]

The telephone company frequency range tests were pretty good, and they certainly served their purpose.  Looking at the reports, though, there was no consideration of equalization such as Don discussed.  The classic applications of equalization were meant to be cancelled for flat throughput, resulting in a better match of signal spectrum characteristics to the transmission or storage medium.  They were not meant to enhance the program audio itself.  Surprisingly, old-time recording engineers said that they did not use any tone control equalization in the old recordings.  Anybody who has worked in sound in the past forty years would be surprised to learn that!  Tone controls were applied by individual listeners to compensate for their room acoustics, speaker characteristics, personal tastes, etc.  Some specialists knew about things like rising response characteristics, but they used microphones like the D-104 to get them, rather than electronic equalization.

In-band equalization strongly influences the audible impact of frequency range filtering.  The kind of rising characteristic Don describes greatly reduces the audible impact of a relatively low high frequency cutoff.  Likewise a broad peak at the low end can reduce the audible impact of a relatively high low frequency rolloff. In recording and broadcasting, a lesser amount of this kind of boost adds clarity and presence to the sound, even when the frequency response extends beyond audibility at both ends of the spectrum.

Like the old-time recording people, hams have generally selected microphones that have desired characteristics.  Some of us go beyond that to further enhance audio with equalization and processing.

[k4kyv]

In-band equalization strongly influences the audible impact of frequency range filtering.  The kind of rising characteristic Don describes greatly reduces the audible impact of a relatively low high frequency cutoff. 

That is very true.  At the flick of a switch I have 3 options for high frequency cutoff:  a 3500~  extremely sharp cutoff brickwall filter; more gradual rolloff starting at 5000~;  or no filter at all.

I rarely run without a filter, but choose between the 3500~ and the 5000~ one depending on band congestion.  Even with the 3500~ cutoff I often get signal reports of "broadcast quality" audio.  The upper midrange boost balances the low end response so that the speech doesn't sound bassy, and partially compensates for the loss of frequencies above 3500~, so that voice articulation is not lost.

The 3500~ cutoff becomes obvious only when I run an A-B comparison.  One other thing that becomes noticeable when running A-B comparison is the distortion caused as the higher frequency audio componets slam into the brick wall at 3500~.  This produces a subtle "rattling" effect to the audio, that most people don't hear unless their attention is called to it.  This effect is due to ringing in the sharp cutoff passive filter and phase shifts at the knee of the curve of the filter.  It is somewhat similar to the distortion caused by mechanical filters in a receiver versus L-C tuned circuits.

I have seen a formula somewhere for balanced sounding (flat response) audio in which a specific mathematical function applied to the low frequency cutoff and high frequency cutoff is a constant.  I forget exactly what it is, and it would take some time to find it.  As I recall, for 3000~ cutoff, a  low frequency cutoff of 300~ is required, but extending the high frequency above this figure allows for more bass response.

[Tom WA3KLR]

The rule of thumb is:  the product of the -3 dB low end TIMES the - 3dB high end equals 400,000 to 450,000, for a balanced tone sound. 

I think it is stated in the Radiotron Designer's handbook, for one.

[k4kyv]

Then the 300-3000 ~ response isn't balanced.  For 3000~ cutoff, the low end would need to extend to 133~.  If the  low end is cut off at 300~, the high end would cut off at 1333.3~, hardly intelligible.

300-3000 does sound like a tin can.

[Tom WA3KLR]

The rule of thumb is not for speech intelligibility, but for a resulting tonal response that sounds balanced in bass and treble for the bandwidth; not having one end deficient.

[Ian VK3KRI]

Co-incidentally, I stumbled across this at work today which may be of some interest (esp. if your interested in operating while under fire at 35,000 ft).
Its one part of a report from 1946 Harvard U Electo-Acoustic Lab sponsered in part by the US Army & Navy Air Forces to better understand communications quality.

If there is interest I can scan the rest of the document which covers testing/design  of microphones, headets  measuring systems etc.

Rest assured that the parts that were classified have been released (Well thats what the stamp on several of the sections says!)

http://www.smoke.com.au/~ic/ham/Audio_Characteristics_of_Communicatiuons_Equipment/Audio_Characteristics_of_Communicatiuons_Equipment_Pt_F.pdf
[File Size ~= 4M]

                                              Ian VK3KRI

[Bacon, WA3WDR]

That's an interesting article.  I never thought about the effects of the reduced air pressure on speech.  I am surprised.  I never flew with the aircraft comm systems I worked on, but I listened to recordings of flight tests done with them, and the audio levels and quality were not an issue.  I would not have known that this effect existed.  Maybe the planes were pressurized.  If I ever work in that field again, I'll take this into account.

It gets really noisy in military aircraft, so of course they use closed-earcup headsets, and they have two watt speaker amps, and the users put in heavy earplugs.  Then the headset speakers blast, and the users can hear.  But there is a lot of mechanical and acoustical coupling of the speakers to the microphone, so DSP echo suppression is needed.  There was interest in ambient-noise reducing headsets, but I didn't see mil-issue units of that type.

Multi-user digital voice systems are tricky when there is time delay associated with the speech digitization.  The stuff I worked with had about ten to fifteen milliseconds of throughput delay.  This made a serious echo sound until the earphones were sealed, and then surprisingly it was not noticeable.

[wavebourn]

How often do you operate your AM stations under a cross-fire?
I hear neither vehicle rumble nor artillery blasts when listening your conversations. 

[Bacon, WA3WDR]

Well you know, I really get a bang out of running AM.

[Tom WA3KLR]

I have heard pilots several times on 121.5 MHz. AM with high voices, but they weren't at high altitude.  It would be safe to say that their gonads were ascended though.

[KB2WIG]

Them pressure suits (flight suits) is tight round da nads...

[wavebourn]

Well you know, I really get a bang out of running AM.

I guess it is really Hi-fi bang...