I have already posted about this a month ago. See post #112.

That is not true. The DAC/ADC support 32 bit reslotution for a few years now. So the person that wrote the delusion comment is clearly not educated enough to understand it

OK so if your explanation is logical you are saying the analog world will never be grater than 24 bit resolution in the digital world ?

I am not sure why but this discussion has turned into a war from a technical point of view. 32 bit float interfaces are not technological marvels and they do not outperform 24 bit interfaces because they are 32 bit. The main reason they are attractive to many is because they are floating point. This offers the ability to map 0dBFS to a lower dBU value without sacrificing dynamic range. So when you record your levels are already near 0dBFS and when you listen to the recordings you don't need to apply a lot of gain. But you still get to keep a lot of headroom in case something unexpected happens. It's also easier to use (especially for beginners) and maybe more consistent since there is no gain knob to adjust...

And sure you could do this with a single 24bit ADC but more is better. Right?

I like to always check the noise floor the old way - by listening. Sometimes there are anomalies and if I have an important sound capture ~ well things can happen.

Gave GS a long rest, sorry for the late reply.

All I'll say is that an audio DAC can't resolve 24 bits at pro audio levels, so it obviously can't resolve 32 bits.

I've found that I can give all the reasons why, and I can include citations that back it up, but the bottom line is that an electrical engineer (or physicist) will already understand why, and someone who is not either will be completely unmoved by the evidence.

So, I'll just say it (you can consider it might be true, or you can think I'm a nutcase making it up): The minimum possible thermal noise in electronics is more than four times the level of the 24th bit. By definition, that means the 24th (and at least the 23rd) are not resolved. 32-bit gives you another factor of 1/256, so how could anyone consider those bits to have meaning? The "32-bit" in 32-bits DACs refers to the math precision, not the resolution of the output.

So, you're right that DACs/ADCs "support 32 bit [resolution]". They absolutely support it. They just don't attain it, but it's physically not possible.

Just something interesting to consider: You're calling the person who wrote the article delusional and uneducated. I don't know what his responsibility is at Mojo Audio (hm, says he's the founder...), and I don't know how good Mojo's DACs are. But consider that they sell their DACs for $4k-$10k, and this guy writes on their blog that "Any company who claims 24-bit resolution from their DAC is simply full of shit...the analog output stage in the world's best DACs are not capable of resolving much more than 20-bits of dynamic range." And that article has been up since May without Mojo taking it down. It sort of seems like the company may believe that's the truth of the matter, doesn't it? (Or is really gutsy with their BS and don't care much about sales...?)

So, either Zwickel, Mojo, and I are right about this, or we're all "delusion[al]" and "clearly not educated enough to understand". If so, I don't know about those guys, but I've managed to eek out a living while making some highly regarded products while being delusional and uneducated. Hey I won't be offended, how do any of us know if we're living a delusion?

Sorry, more wordy than I planned, but I'll add that another thing you can do if you believe we have 32-bit resolution is to find a DAC that specs out with a s/n of 190 dB...or 180...170?...heck, see if you can find 130. Here are the elite-testing DACs (from a much longer list of tested devices) on Audio Science Review; this is SINAD (s/n plus distortion), but to get to this level the distortion is vanishingly small and this is basically a s/n measurement—these include "32-bit" DACs:

Not to beat this point to death, but I had another thought on understanding the base noise level limitation of electronics.

Many know what EIN is—Equivalent Input Noise. It's a measurement popularly used for mic preamps. The test puts a resistor across the input of the preamp, to simulate the output impedance of a typical mic, turns up the gain of the pre, measures the noise at the pre's output (dBu), then subtracts the gain (dBu) of the pre. This results in a dBu value of the effective noise imposed by the pre.

Try to find preamps with an EIN of -130 dBu or lower (larger negative number). You can't, unless the measurement is in error, because that's about the self-noise level of 150 ohm resistor at audio bandwidth. I've had discussions with people who have refused to believe there is an inherent limit to electronics, but this is a fact of physics.

To put into perspective what that means for a 24-bit DAC, the least significant bit amplitude is 2^-24 of full scale (FS), or -145 dBFS. We have to translate between dbFS and dBu (and dBu is defined as an rms ratio...but since it's still a ratio and we know that the gear is spec'd at +4 dBu standard output...). We'll get a little informal and just say that the DAC's FS output gives us the +4 dBu standard output via pre-amplification in the DAC. If the preamp is the quietest we could make, we know it produces about -130 dBu of noise (essentially white in the audio band). That -130 dBu is -134 dB down relative to full scale output of the +4 dBu DAC. (Note that a consumer -10 dBV DAC is nearly 12 dB worse, or about -122 dB down.)

Meanwhile, the lowest bit is -144.5 dBFS, so just over 10 dB quieter than the inherent noise of perfect electronics. In reality, we can't make perfect electronics with no other noise sources, and we probably have higher resistances too (more than -130 dBu—thermal noise depends on temperature, resistance, and bandwidth). That's why the very best DACs on that list from ASR top our at -123.8 SINAD and not -130.

Sorry for going into so much detail and length, but this is why the Mojo article says, "but the analog output stage in the world's best DACs are not capable of resolving much more than 20-bits of dynamic range."

I'm not trying to piss people off, or even change anyone's mind. Just putting this information out for some people with an open mind to consider.

tldr; If you think there is no limit to how small we can get the noise floor, and hence how many bits we can support, ask yourself why we can't make preamps with low enough noise floor to match 24-bits.

80's:
I want an amp that goes to 11.
I wish i could record digital.
2020s:
I want an interface that capture 34 bit
I want a headphone amplifier that can drive a 88000 ohms headphone and can get up to 250db.
I want a 4000$ converter that sounds "musical" and not fatiguing
I wish i could record to tape.

LOL!

And...

1980s:

Young engineer: "Jeesuz H—tape print through is really a menace!"

Senior engineer: "Relax, just store the tapes tails out, the print through junk will be 50 dB below the music, no one will hear it."


2020s:

Senior engineer (the young engineer from the '80s): "Jeesuz H—these people who don't dither their 24-bit masters, there's junk at -140 dB that will ruin the music!"

Hahahahaha

Just got the Platane UP1 interface that is recording at 32-bit float. Sounds better than UAD Apollo, RME Babyface, and Lynx Hilo. It’s on par with a Prism Sound Lyra imo. Makes Scarlett’s and such sound like pure rubbish.

It isn't "recording at 32-bit float". It's providing the sampled data as 32-bit float. For instance, "Routing to phone usb support 32Bit/192KHz".

No offense, but I'm wondering if the "32-bit" spec is making the $129 converter "[sound] better than [a $3599] Lynx Hilo." I guarantee it's not due to added precision.

Whatever helps you sleep at night with an inferior 24-bit interface.

Good one...

32 bit is only important on location fwiw…

Yes. And for the benefit of some I'll point out that for interfaces like the Platane UP1, recently discussed, the 32-bit is in data transport only.

For "location" recording, the 32-bit refers to a different scheme—essentially managing multiple gain levels simultaneously so that capture gain will never be way too little (resulting in excessive noise floor), nor way too much (resulting in clipping). A very important feature when there is no second take, and often no way to know how loud the source is in advance.

But there is never 32-bit conversion in the same sense of 24-bit conversion. It would be like signing your name on the head of a pin with a Sharpie, not physically possible. Nor necessary.

Reason I go 32-bit is because I’m an aliasing freak to my bone. I’m a stretcher and sound designer. 32-bit gives me the QUALITY to do so, whereas the inferior 24-bit sounds like grainy rubbish, which sometimes is necessary, but for my top level productions, absofruitly not.

OK. I myself used 32-bit data for everything. I think that's what you're referring to.

But for the sake of the thread, which is about interfaces that can record 32-bit float, there needs to be the understanding that the interface itself can't record in that resolution, it's physically impossible.

For "location" recording, there can be a ranging scheme that uses 32-bit float to capture multi-resolution, essentially. But the recording resolution is less.

For typically studio recording, there are 24-bit converters that may use 32-bit as the transport. Sometimes, marketing calls a DAC that uses using 32-bit filter math "32-bit".

You never record more than 24-bit. (You never actually record even 24-bit, the physical limits max out somewhere between 20-22 bits.)

You know what I mean big Jean.