> If you are scared of complex numbers for some reason (e.g., if your programming language makes it difficult to work with them) then you may prefer the DHT.
It’s not about that. The DHT cuts memory bandwidth in half. With FFT it’s a complex number, so it’s doubled.
FFT on a real signal can be cut in half too, but it requires doing weird shuffling of the values. (See submission’s blog post on the topic.) It’s unnecessary complexity unless you really need a full FFT for some very specific reason (like the case you mention). Most of the time, people just want to do some DSP like a low pass filter or resizing.
> If you are scared of complex numbers for some reason (e.g., if your programming language makes it difficult to work with them) then you may prefer the DHT.
It’s not about that. The DHT cuts memory bandwidth in half. With FFT it’s a complex number, so it’s doubled.
FFT on a real signal can be cut in half too, but it requires doing weird shuffling of the values. (See submission’s blog post on the topic.) It’s unnecessary complexity unless you really need a full FFT for some very specific reason (like the case you mention). Most of the time, people just want to do some DSP like a low pass filter or resizing.