I’m not against 32 bit, but I think it’s mostly being used by people who lack basic recording skills and who don’t understand its limitations.

Not really sure I understand. The bit rate only really relates to the noise floor of the recording. The loudest thing we can record is 0dBFS, which digitally has a value of 0. With 16bit the quietest thing we can record has a value of 65,535 (16 bits with a value of 1) and with 24 bit this value is 16,777,215 (24 bits with a value of 1). If we try and record something exceptionally quite it could digitally be a number greater than 65,535 (16 bit) or (less likely) greater than 17,777,215 (24 bit). This is most likely going to occur if you set our gain far too low. But if you set your gain too high, you will also get clipping if the recording goes louder than a value of 0 – you cannot represent negative numbers with 16 or 24 bits (unassigned).

All that 32 bit float (not just 32 bit) does is record the loudness in a different format – 1 sign bit (here is part of the magic), 8 exponent bits, and 23 mantissa (fraction) bits. This allows very large negative and positive numbers to be recorded (although not all integers can be precisely represented). The max number that can be presented is 3.4028235 × 10^(38)! So if we set our gain too high instead of clipping, it records a negative number. This of course does not prevent mic distortion (a different problem), if the sound is loud enough to exceed the max SPL of the mic….. And if we set our gain too low, our recording is still there and hasn’t disappeared below noise floor (max value) of a 16 or 24 bit. This latter point was demonstrated by Sound Devices in a example where they set gain to virtually nothing, but still managed to ‘recover’ the recording in post, from what at first appeared to be silence.

Because gain on 32bit float recorders is very forgiving, some recorders don’t really allow control of the gain levels – you just adjust the gain levels in post. This is weird for recordest whose prime concern used to be always getting the gain levels right – now there really is little point!

I would argue it is not really a skills thing – there is more skill in correct mic placement for example. But you could argue that 32 bit float takes away part of the old recording process. I am really not sure what the limitations are though – perhaps battery drain, or possibly large file sizes (although memory is now large and cheap). If I used my MixPre-6II to record in 16 & 24 bit and then in 32 bit float, I think unless I got something hideously wrong with the gain, all recordings would sound the same to the human ear. So to take Peter Jone’s point – there really is no reason the 32 bit float recording should sound any different. CD quality is only 16 bit and the dynamic range (between the loudest and quietest sound) is adequate for most discerning listeners. I remember an experiment where someone subtracted a 16 bit recording from a 24 bit recording and the result was practically silence – I would kind of argue that 24 bit is therefore being used by people who do not really understand what it means and in most cases it is unneeded. However, I would argue that 32 bit float adds utility. I also think that for bird recording 32 bit float is very useful – if I was recording music, I can ask the musician for a sound check to adjust recording levels, so that I am hitting the top of my ballistics but not going over. Birds are not so co-operative, so getting it right first time and without a sound check is important!

Finally I would say that I don’t think 32bit float is that novel anymore – The MixPre 6II was launched over 5 years ago, and most new recorders on the market now have 32bit float capability (including devices by Tascam, Sound Devices, Zoom, Tentacle, Rode etc.). Although if you are in the market for true professional stuff then the Nagra, Sonosax and Zaxcom are all non-float recorders, but this is the world of £5k plus devices, which do a lot of professional stuff – and changing from non-float to float is not an easy switch, so I imagine it will be a bit like Hassablad not jumping into digital photography – but look at them now!