Default Sample Format; why higher than 16bit?

Dear all,

I’ve connected my vintage record player to a Pro-Ject phono amp. This phono amp has a build-in 16bit 48kHz ADC with USB connection (and gain control).
So I connected the amp via USB to my mac, and via Go-Utilities-AudioMidiSetup I changed the settings to 16bit / 48kHz (iso the default 44.1 kHz).

Started up Audacity, and in Preferences-Quality I selected 48000 Hz as Default Sample Rate and 32bit as Default Sample Format.
So far so good, and recording LP’s is without any issue.

The gain control on the phono amp is actually that good, that it’s quite easy to set the maximum input level around -3dB, even close to 0dB if I want to.
No clipping, and at the end no need to Normalize the file. At least I see no reason for that (am I missing an important step here?).
I store my files as AIFF, 16bit/48kHz.

My question is this; if I do not apply any filtering -besides a fade out at the end of the recording- is there any need to have Default Sample Format on anything higher than 16bit?

Thanks for the help,

The gain control on the phono amp is actually that good

You may have some extraordinarily well-behaved records. I couldn’t get my collection to do that. I had to reset each record and sometimes during the record it would wander as the theatrical expression picked up.

My system is of good quality and relentlessly manual, so something in your system may be trying to automatically “help” you set levels.

You should be saving your work as uncompressed WAV or AIFF. You can always work from there down to the other compressed formats, MP3, M4A, etc. for posting on-line or your personal music player, but you can’t come back up again. Audacity default is WAV because WAV will easily open up anywhere. AIFF will only open on Macs and some other special-purpose software.

Audacity always works internally at 32-floating, so you’re going to get a conversion or two no matter what you do. It does that to cope with effects and filters that temporarily increase the volume of work beyond 16-bit overload. At 32-floating, you can recover from this damage.

48000 is the video sample rate. 44100 is the Audio CD sample rate. The only different is slight increase in accuracy above 16KHz.


No, not really, though I know that with my record collection I would need to play the record at least twice for each recording (once to set the level and once for the actual recording) and in the case of some classical records I’d probably need to play the record more than twice to ensure that the highest peak was at the desired level.
Personally I find it far more convenient to set the level “about right” (allowing say 6 dB of headroom) and Normalize before exporting. Also, few of my (old) records are completely scratch free, so I’d prefer to remove or at least reduce some of the clicks. It is when applying multiple processes that the benefit of 32 bit float format really starts to count.

On the other hand, working in a higher bit format will do no harm. It can be likened to using a printer that can handle 2000000 dpi when you only really need 600 dpi.

Thanks for all the help, and I will use 32bit from now on.
I was thinking: my ADC outputs 16bit, and my AIFF file will be 16bit. So why risk upconversion to 24 or 32 bit and afterwards a downconversion to 16bit. I thought the less conversions the better.

Btw, most of my records are in a perfect state (I also clean them in a professional vinyl washing machine). That helps reducing the clicks without the need for a software filter.
And before the actual recording, I also play the whole record to make sure the peaks stay around -3dB. The gain control on the ADC makes it very easy to do so (easy to finetune), much easier than via the software controls.
I thought this way there is no need for Normalizing, at least not to correct the volume, only to correct the offset.

Is this in principle not a better approach than to record at around -6dB and Normalize?

Converting from 16 bit to 32 bit is totally lossless. Every 16 bit value has an exact representation in 32 bit format, so the conversion from 16 bit to 32 bit is absolutely perfect.
In theory, a fade out done in 32 bit float format, then converted to 16 bit will be better than a fade out done in 16 bit. In practice the difference is likely to be so small as to be unnoticeable.

Discussions about differences between 16, 24 and 32 bit float tend to revolve around theoretical differences. In most situations the differences are negligible.

I can give a couple of practical examples to illustrate the pros and cons if you are interested?

Thank you for the explanation steve, it helps me getting the right settings in Audacity for the best results.
In Preferences-Quality I selected 48000 Hz as Default Sample Rate and 32bit as Default Sample Format.

Again, thanks to you all for your advise and quick response.