Since those are digital filters there could be gain compensating
factors, that are over compensating, or different styles of filters and
the filter coefficients don't match the gain from one bandwidth and
shape factor to another. One of the nagging details of DSP filters is
that you have to keep track of levels in accumulators so as to not go
over the range of available numbers, that's most critical in a 16 bit
integer DSP, fairly critical in 32 bit integer DSP and hardly a problem
in a 32 bit floating point DSP. Trouble is the 16 bit integer DSP is
cheapest and fastest, the 32 bit integer DSP can be low cost and very
fast, while the 32 bit floating point DSP is inherently most expensive
73, Jerry, K0CQ
On 7/17/2010 12:41 PM, Jay Lyle wrote:
> This seems strange to me:
> 1) Right now, noise level with 6k/2600 is 3-4 s-units
> 2) If I go down a notch to 2.5k/2500 noise GOES UP to 4-5
> Is this observation is consistent across all bands (blankers/notch and such
> You'd think it would GO DOWN with a narrower filter combo.
> Going down to narrower settings, I get the expected results.
> 2.5k/600 shows 4 s units
> 500/500 GOES DOWN to 2
> I believe radios will have different gain settings depending on different
> filters to make up for the insertion loss.
> Anyone else see this effect?
> TenTec mailing list
TenTec mailing list