On Thu, 31 Jul 1997 10:22:00 +0100 Oyvind Kaurstad
<oyvind.kaurstad@NOFAC.ABB.NO> writes:
>>Oyvind Kaurstad writes:
>>> I'm planning on designing a "channels in use" indicator
>>> for R/C purposes. I know there are commercial units for
>>> sale, but they are expensive.=3D20
>
>>[snip]
>
>>> The hard part is obviously the RF design, and I wonder if anyone
>>> has any cheap and clever ideas on this.
>>>
>>> I live in Norway, and the RF frequency in question is 35.xxx MHz.
>>>
>>> The channel spacing is 10 kHz.
>
>>Everyone else who responded to this seems to think you were asking
>how
>>to build a superhet receiver. I sort of thought you were asking how
>to
>>cheaply build 40 or 80 receivers. If that is what you were asking,
>>I'll give you my ideas on how to do it. I'll assume you have 80
>>channels like in the US. To do this job cheaply you must reduce the
>>hardware. That is, build one single receiver that mixes the entire
>>35.XXX band down to baseband. You can decide whether it is best to
>use
>>a heterodyne or superhet receiver. Once the signal is at baseband,
>low
>>pass filter it to 80*10=3D800 kHz, sample and hold and 8-bit digitize
>>with at least 1600 kHz sampling. Store, say, 512 samples in your
>>micro's memory and then FFT it.

The major problem with this is the dynamic range of an 8-bit converter,
about 30dB.  Since the device is to be used at the flying field there
will be many nearby transmitters but still a need to detect far-away
ones.  A strong local signal will dominate the ADC and make signals more
than 30 dB weaker undetectable.  A MHz-sampling, 12 or 16 bit ADC which
would start to address this problem is still an expensive proposition.

This type of receiver is a good idea when several signals of
approximately equal strength are involved, or there is a need to rapidly
identify newly-appearing signals.

To me, the sweeping superheterodyne seems a more viable solution for R/C
since the band is narrow enough to use a fixed-tuned front end and a
simple VCO.

>>You could probably choose a window
>>function that gives a broad mainlobe and low sidelobes because you
>are
>>not really interested in resolution but in detection. Use a threshold

[...]


>I do know how FFT works, and I agree, it would not be recommendable to
>try doing it on a PIC. At least not the '84.

Thie first problem is the PIC's (any model) lack of sufficient RAM to
store the original samples, let alone the transformed results.