Hi, They say that a receivers sensitivity is proportionate to it's bandwidth. The narrower the crystal IF filter the more sensitive the receiver becomes. I'm talking about a traditional radio and as an example let's say I'm just trying to detect a CW with a extremely low bit rate. They say you can gain about 10 dB of sensitivity from going from a 12.5kHz wide IF analog filter to a analog roofing filter (500Hz wide). So if I sample the IF with the 12.5kHz wide filter in and apply a software filter with the same bandwidth and shape factor as the 500Hz roofing filter, will my software radio gain similar over all sensitivity? Simply put, can a narrow software filter replace a narrow hardware filter to produce the same overall sensitivity of a radio? Thanks, Thomas Magma
receiver sensitivity, harware vs. software filter?
Started by ●May 3, 2007
Reply by ●May 3, 20072007-05-03
Thomas Magma wrote:> Hi, > > They say that a receivers sensitivity is proportionate to it's bandwidth. > The narrower the crystal IF filter the more sensitive the receiver becomes. > I'm talking about a traditional radio and as an example let's say I'm just > trying to detect a CW with a extremely low bit rate. They say you can gain > about 10 dB of sensitivity from going from a 12.5kHz wide IF analog filter > to a analog roofing filter (500Hz wide). So if I sample the IF with the > 12.5kHz wide filter in and apply a software filter with the same bandwidth > and shape factor as the 500Hz roofing filter, will my software radio gain > similar over all sensitivity? > > Simply put, can a narrow software filter replace a narrow hardware filter to > produce the same overall sensitivity of a radio?Not quite. The bottleneck is the dynamic range of the receiver. This parameter is limited by the amount of garbage in the bandwidth of the first filter. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●May 3, 20072007-05-03
Thomas Magma wrote:> Hi, > > They say that a receivers sensitivity is proportionate to it's bandwidth. > The narrower the crystal IF filter the more sensitive the receiver becomes. > I'm talking about a traditional radio and as an example let's say I'm just > trying to detect a CW with a extremely low bit rate. They say you can gain > about 10 dB of sensitivity from going from a 12.5kHz wide IF analog filter > to a analog roofing filter (500Hz wide). So if I sample the IF with the > 12.5kHz wide filter in and apply a software filter with the same bandwidth > and shape factor as the 500Hz roofing filter, will my software radio gain > similar over all sensitivity? > > Simply put, can a narrow software filter replace a narrow hardware filter to > produce the same overall sensitivity of a radio?Not quite. The bottleneck is the dynamic range of the receiver. This parameter is limited by the amount of garbage in the bandwidth of the first filter. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●May 3, 20072007-05-03
Vladimir Vassilevsky wrote: ...> The bottleneck is the dynamic range of the receiver. This parameter is > limited by the amount of garbage in the bandwidth of the first filter.It's too bad the analog front end can't just add more bits to get more dynamic range. Designs would be so easy if only we didn't have to make them work! Jerry -- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by ●May 3, 20072007-05-03
"> Not quite.> The bottleneck is the dynamic range of the receiver. This parameter is > limited by the amount of garbage in the bandwidth of the first filter. > > Vladimir VassilevskyIt is a fixed frequency receiver with a low insertion loss 5 MHz wide RF SAW filter, a 10 MHz wide first IF SAW filter and a 12.5 kHz wide second IF crystal filter. My options are to remove the crystal filter and replace it with a 500 Hz wide roofing filter or leave the crystal filter in and digitize and implement a 500 Hz wide (or narrower) software filter. I'm just trying to determine the pros and cons each approach, especially the sensitivity. Thomas
Reply by ●May 3, 20072007-05-03
On May 3, 5:11 pm, "Thomas Magma" <somewh...@overtherainbow.com> wrote:> "> Not quite. > > > The bottleneck is the dynamic range of the receiver. This parameter is > > limited by the amount of garbage in the bandwidth of the first filter. > > > Vladimir Vassilevsky > > It is a fixed frequency receiver with a low insertion loss 5 MHz wide RF SAW > filter, a 10 MHz wide first IF SAW filter and a 12.5 kHz wide second IF > crystal filter. My options are to remove the crystal filter and replace it > with a 500 Hz wide roofing filter or leave the crystal filter in and > digitize and implement a 500 Hz wide (or narrower) software filter. I'm just > trying to determine the pros and cons each approach, especially the > sensitivity. > > ThomasIt depends on how you intend to test the sensitivity. If you connect a signal generator to the receiver and drop the level down until a certain output SINAD or BER is acheived, then the software solution should be able to match the hardware one. If you repeat the test with an adjacent channel signal 150 dB stronger, the software will not be able to match the hardware.
Reply by ●May 3, 20072007-05-03
Thomas Magma wrote:>>The bottleneck is the dynamic range of the receiver. This parameter is >>limited by the amount of garbage in the bandwidth of the first filter. >> > It is a fixed frequency receiver with a low insertion loss 5 MHz wide RF SAW > filter, a 10 MHz wide first IF SAW filter and a 12.5 kHz wide second IF > crystal filter. My options are to remove the crystal filter and replace it > with a 500 Hz wide roofing filter or leave the crystal filter in and > digitize and implement a 500 Hz wide (or narrower) software filter. I'm just > trying to determine the pros and cons each approach, especially the > sensitivity.It depends. You have to run the numbers. The key parameters are IP3, noise figure, the required attenuation of the adjacent channel, and how much money are you willing to spend. BTW, are you sure the frequency is accurate enough to fall into the 500Hz bandwidth? That will require the clock sources with the tolerance of 0.1ppm or so. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●May 3, 20072007-05-03
> It depends. You have to run the numbers. The key parameters are IP3, noise > figure, the required attenuation of the adjacent channel, and how much > money are you willing to spend. >Well there is not really any adjacent channel to speak of, and the noise figure is pretty much set by the saw filter in front of the first LNA, so there is not much I can do about that. The answer to how much money I'm willing to spend would be 'not much'. I'm pretty much a minimalist when it comes to designs and that is why if I can kick that expensive (and bulky) roofing filter out of the design and replace it with a software filter I'm all for it.> BTW, are you sure the frequency is accurate enough to fall into the 500Hz > bandwidth? That will require the clock sources with the tolerance of > 0.1ppm or so. >If the analog roofing filter were to remain, the receiver would have scanned and centered on the signal, but if I go software filter then I can FFT first before applying the narrow software filter, so I guess there is that advantage. Thomas
Reply by ●May 3, 20072007-05-03
On May 3, 4:22 pm, "Thomas Magma" <somewh...@overtherainbow.com> wrote:> > They say that a receivers sensitivity is proportionate to it's bandwidth.that's what i heard.> The narrower the crystal IF filter the more sensitive the receiver becomes.now which is it? proportional and inverse-proportional is not the same.> I'm talking about a traditional radio and as an example let's say I'm just > trying to detect a CW with a extremely low bit rate. They say you can gain > about 10 dB of sensitivity from going from a 12.5kHz wide IF analog filter > to a analog roofing filter (500Hz wide).okay, maybe S/N increases as the BW gets smaller to a limit that has something to do with the bit rate. if the "roofing" filter (not completely sure what the terminology is) is an LPF of 500 Hz you can get something like 1000 bit/s. if it's a 500 Hz wide BPF operating on a complex IF signal, it would be about the same 1 Kb/s.> So if I sample the IF with the > 12.5kHz wide filter in and apply a software filter with the same bandwidth > and shape factor as the 500Hz roofing filter, will my software radio gain > similar over all sensitivity?ideally, if your bit rate is limited to 1 Kb/s, and your digital filter can be a really sharp brick-wall (if you don't mind the delay and you can throw an FIR with as many taps as you need at the problem of designing the 500 Hz filter (whatever "roofing" means ... maybe you mean "ceiling" which sorta has the same meaning as "brick wall" in this context).> Simply put, can a narrow software filter replace a narrow hardware filter to > produce the same overall sensitivity of a radio?yes, but the best word to use is *selectivity* instead "sensitivity". but if "sensitivy" is S/N and you have a fixed and limited bit rate, increasing selectivity increases sensitivity. and you can make *very* selective digital filters, if you're willing to put enough taps in it, or maybe an IIR LPF like a zillionth-order Butterworth, either way if resources is not the issue then delay is what kills you. r b-j
Reply by ●May 3, 20072007-05-03
On 3 May 2007 15:06:38 -0700, sampson164@gmail.com wrote:>On May 3, 5:11 pm, "Thomas Magma" <somewh...@overtherainbow.com> >wrote: >> "> Not quite. >> >> > The bottleneck is the dynamic range of the receiver. This parameter is >> > limited by the amount of garbage in the bandwidth of the first filter. >> >> > Vladimir Vassilevsky >> >> It is a fixed frequency receiver with a low insertion loss 5 MHz wide RF SAW >> filter, a 10 MHz wide first IF SAW filter and a 12.5 kHz wide second IF >> crystal filter. My options are to remove the crystal filter and replace it >> with a 500 Hz wide roofing filter or leave the crystal filter in and >> digitize and implement a 500 Hz wide (or narrower) software filter. I'm just >> trying to determine the pros and cons each approach, especially the >> sensitivity. >> >> Thomas > >It depends on how you intend to test the sensitivity. If you connect a >signal generator to the receiver and drop the level down until a >certain output SINAD or BER is acheived, then the software solution >should be able to match the hardware one. > >If you repeat the test with an adjacent channel signal 150 dB >stronger, the software will not be able to match the hardware.I agree with this assessment. The filtering trick you propose will provide a deterministic amount of processing gain, but that can't make up for distortion levels that happen prior to that. As Vladimir mentioned, often what happens is the receiver drives the AGC up on the small signal, but if there isn't enough dynamic range to handle the composite power (i.e., the desired plus any adjacent signal energy in the input bandwidth of the LNA or the rest of the chain), and the AFE is driven into saturation. So the whole chain needs to be able to handle the desired behavior once you've pulled the narrow signal out of the noise. Generally, if you're -only- noise limited in the AFE then taking more processing gain through filtering is a good way to increase sensitivity. If the radio was designed properly, i.e., not any better than it really needs to be, then that's seldom the case. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
