It sounds to me that we are talking about three integrated mechanisms that create Ultrasound as a usable sense. As a sensory mechanism, it seems that you need the ability to interpret sound waves as a visual medium. This doesn't require being able to transmit them, and this doesn't require ultrasonic frequencies per se. Just the ability to focus on a single frequency and use the delays and other acoustical clues to build a mental model of the physical environment you are in. This seems like it might require Select Sound Filter to pick out a given frequency, and then the actual Improved Sense: Echolocation Adept Power to make use of the filtered frequency band. Sound Filter is allowed as an adept Improved Sense, so we are good here. Since we have examples of people today that can use echolocation, it could even potentially be done without Select Sound Filter. That would have to be decided on a group by group basis.
If you want to be able to work with actual Ultrasound, then you need Audio Enhancement, which can't be used as an Improved Sense as it has a wireless bonus, and might even not expand the hearing range enough to include the Ultrasonic Frequencies used by Ultrasound sensors - it simply states that it allows hearing of sounds above and below normal metahuman hearing range. I don't see any reason not to allow it, but that would be a house rule.
And lastly, you need to be able to generate the sound waves you are using for echolocation. This can be as simple as clicking your tongue (as people today do) or using Ultrasound Transmitters (or some magical voice adept power that I don't think exists in 5e yet). If you opt to use sound waves in the normal human hearing range (if you just take echolocation by itself for .25 PP) then you can use see fine depending on background noise, but everyone else can hear whatever sound source you are using (so if you are using clicks, they can locate you as easily as they can localize any other sound they can hear). With select sound filter, you can reduce or eliminate penalties due to background noise. With Audio Enhancement you can echolocate with sounds that others won't hear, allowing more stealth options.
So, I don't see any real reason to allow echolocation as a .25 PP adept Improved Sense, but with just that you have to use metahuman detectable frequencies and use other means to generate the sounds that you use. You'd be effectively like the blind guys today that use echolocation, though perhaps with better fidelity. It remains a passive only sense just like the other possibilities, and you might even incur penalties due to the lower frequency sound waves not allowing finer details to be noticed, like textures - I'm not an audio expert, but it seems to me that the detectable size limit for audio perception would probably follow something along the lines of 1/2 the wavelength of the sound wave used (
https://www.nde-ed.org/EducationResources/CommunityCollege/Ultrasonics/Physics/defectdetect.htm), and that doesn't account for scatter over distance, so progressively larger details would become undetectable the further the waves travelled from the generator to you, making the use of remote sound generators incur further penalties.
Exact rules to implement penalties for using lower frequencies and remote sound generators would be handy, but I suspect they could be fairly complex to construct considering the number of total possibilities for implementing echolocation. Bats use between 12khz up to 100khz and higher. Underwater echolocation tends toward the upper end of that spectrum. Human hearing is 20hz to 20khz, meaning that decent echolocation should be possible within the range of normal human hearing. But echolocation can be done using frequency modulation (using multiple frequencies) which provides more detail, but reduced range, or constant frequency, which has better range but poorer resolution and also appears to be good for determining the target's speed due to doppler-shift.
I spend way too much time on Wikipedia. Point being, I couldn't make any rules that might even remotely model real world scenarios, so mostly I'd just apply say a -2 penalty to perception when not using ultrasound for echolocation and another -2 when using a remote audio source, in addition to any environmental mods due to background noise the GM deems appropriate. This means that just taking Improved Sense: Ultrasound (and possibly Audio Enhancement) gives you the ability to hear ultrasound but not echolocate. Improved Sense: echolocation gives you the ability to use sound to echolocate, but not to hear nor produce ultrasound (thus at a net -4 penalty to perception tests). Spending .5 PP for both gets you passive Ultrasound echolocation that would suffer a -2 penalty due to the remote sound generators - though if you carried the generator on your person it could count as local, but you then broadcast your position to other Ultrasound detectors. And so on. I'd also halve the range for not using Ultrasound.
Sorry for the wordiness.