Mike_G_D; Thank you for the detailed explanation. I have not used a Uniden with a NFM channel, so I don't know about them. But I have used all of the GREs and guessed they did not do it correctly. I can see them adding a filter then implementing their "audio boost" on the channel. Uh huh.
I can't say I've monitored a single signal that changes when I apply FM vs NFM though. BUT, if the signal is narrow, then the filter would have nothing to remove. And for whatever reason there's no audio boost.
Not sure if you intended it but your post is showing up as in all italics for me - just FYI.
Anywho, I'm not sure I get what you're saying in your second paragraph; I can step you through a procedure to prove (or disprove) the actions (or inactions) of an IF filter change (or not as the case may be) if you like. This is without test equipment; if you have access to proper lab equipment then it's simpler and more exact, of course. I can do it either way but since I lack access to said equipment now I can only use the former (not using test equipment) method in practice at home. I've gone through it many times on these forums but I will again if you so desire or we can do so via PM.
Insofar as what the filter will or will not do versus the bandwidth of an FM signal - regardless of the deviation it will certainly have an effect. We "decide" that a certain bandwidth is acceptable for the proper reception of an FM signal at some chosen deviation; technically or rather theoretically, the bandwidth of the transmission goes on for quite some spacing (theoretically infinite) outside of Fcenter (ideally symmetrically about the center). But those sidebands become less and less significant the further out from Fc that you go and can be filtered out without serious degradation to the received and demodulated audio. So, for narrowband FM with deviations close to or under the maximum audio frequency response (FM deviation is directly related to loudness, of course, but is also tied to audio frequency range via the modulation index equation as deltaF/highest frequency modulating signal component) we can use receiver IF bandwidths in the 10KHz to 25KHz range. Generally, we take a value of much less than 1 as being narrowband FM while values much greater than 1 are considered wideband FM (like analog TV audio and analog broadcast FM). Anyway, the idea in terms of a receiver IF filter is that we determine at some point beyond which, bandwidth-wise around Fc, we wish the filter to "cut off" or seriously attenuate any energy. This is determined by the required bandwidth for proper reception (for analog systems Carson's rule works ok so we can use that as a general guide for what filter bandwidth to design for; about 11KHz for 2.5KHz deviated systems and about 16KHz for 5KHz deviated systems) and by the required channelizing scheme being used (for example, spacings of 15KHz or 7.5KHz, etc.). Many times it's, as in all things engineering, a compromise. In any case, whatever filter is being used in a receiver it will have an effect on the received signals both desired and undesired.
Just because all signals present in the RF passband prior to the IF stage of a FM receiver are all "narrow" in the sense that they are using the newer "narrower" narrowband FM deviation of +/-2.5KHz does not mean that they will not have complex undesired affects to the desired signal. Restricting ourselves to only adjacent and alternate channel effects (what we mean by receiver "selectivity" in the lab), let's say your receiver uses an IF filter with 6dB down at about +/-18KHz and 50dB down at +/-50KHz. With two or more 2.5KHz deviated signals located within 7.5KHz of each other as in the North American civilian VHF High band channelizing scheme and with sufficient power you will certainly experience issues trying to separate those signals. Even if they are spaced 15Khz from each other, with such a filter, strong or moderate signals even at 2.5KHz deviation can have an effect on each other. The effects would certainly be worse if one or more of the undesired signals were using wider deviation, of course, but with sufficient strength even the 2.5KHz deviated signals will cause problems even at 15KHz away from the desired. Now, if you change that filter to a narrower one with, say 6dB down at 11KHz and 60dB down at 18KHz you will, of course, see a difference, again, in either deviation case with moderate to strong levels of adjacent channel activity.
I've seen two ways to handle the IF filtering of 2.5KHz deviated analog FM signals versus 5KHz deviated signals - using one filter and using two (or somehow modifying the response of one filter as some claim is the case - for the sake of simplicity for this discussion we'll call that the same as effectively "two filters"). In other words, you use one typical IF filter for all "narrow" modulation modes and call it a day or you reduce your IF bandwidth for special "super narrow" modes like the new "narrower narrowband FM". I own two examples of the former method and three examples of the latter. The two of the former are my Uniden BCT15 and my Wouxun HT. The three of the latter are my three GRE made scanners, two Radio Shack desktop/mobile Pro-197's and one GRE PSR500 handheld. All three of the later units work similarly as one might expect - switching in the "NFM" filter reduces the IF bandwidth. The two of the former, however, while operating as one might expect (no IF bandwidth change) do behave differently relative to each other in the sense that the Wouxun HT has a more effective IF filter with, apparently, better stop band attenuation and maybe steeper skirts than the Uniden BCT15 (I would really need to put them on the bench with proper test equipment to fully characterize this, of course). In operation, it is obvious that the HT simply has a better IF filter than the Uniden scanner. So, while they may both compromise in the same way in the sense that they both only use one IF filter response to accommodate both 2.5KHz and 5KHz deviated analog FM signals the fact that the HT has a better overall filter means it works much better when dealing with closely spaced signals of either deviation than the Uniden. Again, I have proven this to myself in operation and can absolutely say that the Uniden has severe adjacent channel issues when confronted by multiple moderate to strong analog FM signals spaced 15KHz or less from each other regardless of deviation. The Wouxun HT simply works better because its one IF filter is better. The GRE made units in standard regular narrowband FM (what many now call "wideband" - ugh!) meaning the +/-5KHz deviation mode react similarly to the Uniden but when switched to "NFM" mode the IF passband is sharply reduced and is very evident when dealing with adjacent and alternate channel energy. When not in the presence of any such undesired signals you will likely not be able to tell any major difference in demodulated audio between the two modes as even the narrower setting seems wide enough for passage of most of the needed information bandwidth of a "wide" 5KHz deviated signal. Thus, even when listening to "wide" 5KHz deviated analog FM signals (such as Marine and amateur signals and those Part 90 units not yet changed over either with or without official FCC waivers) using the "NFM" mode can help when dealing with closely spaced interfering signals - a side benefit, of sorts, due to the GRE "laziness" of not changing discriminator response or otherwise compensating for the 2.5KHz deviation signal audio.
In short - you will likely not hear any change (or hear very little change) in audio in GRE made units when listening to a narrowband (either 2.5KHz or 5KHz deviation) analog FM signal when switching between NFM and FM modes but you will notice improvements in adjacent channel selectivity when engaging NFM mode over the FM mode. In Uniden units just the opposite will likely be experienced - you will notice audio changes when changing between FM and FMN modes but will see, or rather hear, no difference in adjacent channel selectivity; when no adjacent channel activity of sufficient strength is present, of course, it would be difficult to impossible to tell. At least, for the units I own, this has been my experience.
-Mike