Transferring Cassettes to PC

[Kerb26]

I’m going to transfer some home recordings from cassettes to PC.

I found a couple of articles from SOS that suggested using 16 bit as this would be sufficient for the dynamic range of cassettes. The articles are quite old, so I wanted to ask if this is still the best approach?

The recordings that I’m transferring are very inconsistent in level which makes it difficult to set the gain properly. Obviously this would be easier in 24 bit, but the file sizes would be bigger.

I’ve also heard some people say that you should use 24 bit if you plan to do any processing, but I’m not sure if that’s correct?

[Tomás Mulcahy]

16 bit is more than sufficient for cassette. Let's prove that with maths. Rule of thumb, each bit improves SNR by 6dB. Absolute best case for cassette with a top notch deck and NR engaged is around 60dB. That's 10 bits. And it's highly unlikely that home recordings will have SNR as good as that. So again, 16 bits is more than enough.

[Hugh Robjohns]

Yes, 16 bits would be sufficient, but you'd be pretty challenged to find a 16 bit converter or interface... they are all 24 bits!

And if your cassette transfer is going into the computer at 24 bits, its a bit pointless reducing to 16 bits, especially since any processing you apply will be done at 32bit floating point or higher.

Yes, the file size will be a bit larger, but that's not going to trouble your computer or its storage media.

[ef37a]

Well Hugh, the Berry UCA 202 is 16 bits and, if the OP does not want to spend out much for a one off job is totally good enough for cassette. I have had 3 and all give a noise floor around -80dB fs, i.e. not quite as good as 16 bit theory.

Then, most DAWs give the choice of 16 or 24 bits? Samplitude certainly does. But yes, better to run at 24 bits and keep levels around -18dBfs. If the chap is concerned about file size, once 'processed' they can easily be 'saved as' 16 bit files.

Dave.

[Kerb26]

And if your cassette transfer is going into the computer at 24 bits, its a bit pointless reducing to 16 bits, especially since any processing you apply will be done at 32bit floating point or higher.

My interface is 24 bit. I assumed it would convert at whatever bit depth I set in my DAW. But it sounds like you’re saying that the interface would capture the audio at its native bit depth (24 bit) then my DAW would reduce it to 16 bit?

If so, presumably this would introduce quantization noise?

[Hugh Robjohns]

My interface is 24 bit.

Yes, most are. You've got to go out of your way to find some old 16 bit nonsense these days...

I assumed it would convert at whatever bit depth I set in my DAW. But it sounds like you’re saying that the interface would capture the audio at its native bit depth (24 bit) then my DAW would reduce it to 16 bit?

Yes.

If so, presumably this would introduce quantization noise?

Only if the wordlength is reduced by truncation rather than dithering.

But this isn't worth worrying about. Just set the level with a generous headroom margin so you don't have to worry about potential clipping. Average around -20dBfs with peaks to -10dBfs is a good target. Record with 24 bits. Once the transfer is complete you can optimise the levels to suit your chosen destination's requirements.

[ef37a]

Actually there was a recent SoS review of a couple of M-Audio interfaces that were 16 bit, budget jobs mind.

The tape noise is going to be way higher than any artifacts.

Dave.

[Kerb26]

Thanks to everyone who commented.

[James Perrett]

Just remember that 16 or 24 bits should be the least of your worries when transferring cassettes. Mechanically aligning the playback machine to match the tape is far more important as, once you've transferred a tape with the wrong alignment, there is no way of retrieving the information that you've lost (apart from transferring again with the correct alignment).

You may also need to think about noise reduction - although I usually transfer without noise reduction and then do the noise reduction subsequently in the appropriate hardware or software so that I can fine tune the levels and eq.

[Tim Gillett]

16 bit is more than sufficient for cassette. Let's prove that with maths. Rule of thumb, each bit improves SNR by 6dB. Absolute best case for cassette with a top notch deck and NR engaged is around 60dB. That's 10 bits. And it's highly unlikely that home recordings will have SNR as good as that. So again, 16 bits is more than enough.

Some cassette decks incorporated DBXII NR which from memory was around 80db S/N requiring more care when setting transfer levels.

[ef37a]

16 bit is more than sufficient for cassette. Let's prove that with maths. Rule of thumb, each bit improves SNR by 6dB. Absolute best case for cassette with a top notch deck and NR engaged is around 60dB. That's 10 bits. And it's highly unlikely that home recordings will have SNR as good as that. So again, 16 bits is more than enough.

Some cassette decks incorporated DBXII NR which from memory was around 80db S/N requiring more care when setting transfer levels.

I agree Tim, I have a Sony Dolby C machine that can copy CD (TDK SA) such that it is hard to tell from the original. However, OP is dubbing "home recordings" it is very unlikely that the dynamic range of the original material is anything like 80dB.

'GIGO'! (wtgr)

Dave.

[Nazard]

Rather surprisingly, some cassette decks were capable of sounding extremely good. For example, I have a Tandberg TCD 440A, which was straight from Tandberg, (then in Leeds), and the SA90/SA90X tapes, still sound terrific.