[BC] Analog AM bandwidth: tail wagging the dog?

[Robert Orban]

At 02:08 PM 10/4/2007, you wrote:
>Bob:
>
>
>I have had great success in the analog world with a 5 kHz bessel 
>filter with extra poles added so that the response is severely 
>attenuated at 7 or 7.5 kHz and upward. It seems no matter what 
>response is chosen there are always issues with having audio 
>response right up to the sharp cut off of the final filter. Maybe 
>you and/or Frank might consider an adjustable prior to the multi 
>band section in some future processor. The parametric EQ in the 9400 
>can accomplish some of this but a dedicated LP filter would be much better.

I'm not sure if you're talking about the parametric EQ section or the 
parametric lowpass filter shape in the 9400 (and 9300), as these are 
separate and distinct features. The lowpass filters in these products 
are FIR. The "0.1 dB" is basically a classic Parks/McClellan 
equal-ripple passband FIR lowpass filter. The "-3 dB" and "-6 dB" 
versions shape the frequency response of the FIR filter around its 
cutoff frequency to approximate a transitional Bessel/Butterworth 
filter out to the -6 dB and -12 dB points on the lowpass filter 
frequency response curve. Eventually, the filter transition band 
response curve has to get very steep in order to meet spec for 
transmission channel spectral protection, but the "nose" of the 
filter becomes progressively smoother as one adjusts the "shape" 
parameter from 0.1 dB to 6 dB.  So the 9300 and 9400 already have a 
feature that's very close to the one you are describing -- the "-6 
dB" version is probably closest. (Figure 2-8 in the 9400 manual shows 
the frequency response curves for the parametric lowpass filter at 5 kHz.)

Bob Orban





>Robert Orban wrote:
>>At 06:19 PM 10/3/2007, you wrote:
>>
>>>Bob Orban wrote:
>>> > I believe that a significant reason that the final preferred
>>> > bandwidth was 7 kHz was that broadcasting wider bandwidths puts more
>>> > stress on the audio processor. If the processor clips material
>>> > between 7 and 10 kHz (to which a vast majority of AM radios are
>>> > essentially deaf), the resulting clipping distortion products *will*
>>> > end up in the 0 to 3 kHz frequency and *will* be heard on typical
>>> > radios.
>>>
>>>That's all the more reason why audio above 6 kHz or so should be given its
>>>own band in the processor's multiband compressor/limiter structure, like
>>>your 9100 did, so that most of the HF gain control can be achieved before
>>>the audio hits the clipper(s).
>>
>>This comment misses the point. If the transmission includes energy 
>>between 7 and 10 kHz, it takes up extra "room" in the modulation 
>>envelope. Getting a 0 - 10 kHz bandwidth transmission to sound as 
>>loud as a 0 - 7 kHz transmission on an average radio requires the 
>>audio to be clipped and limited harder, regardless of the details 
>>of how it is done in the audio processor.
>>bob Orban
>_______________________________________________
>
>The BROADCAST [BC] list is sponsored by SystemsStore On-Line Sales
>Cable-Connectors-Blocks-Racks-Test Gear-Tools-Lots More + Now Barix too!
>www.SystemsStore.com       Tel: 407-656-3719    Sales at SystemsStore.com
>
>