Improving the quality of wav files from cassette

Hello all! I’m on Windows 10 and using Audacity 2.1.3 (.exe installer)

I’ve tried some sample cassette-to-digital recordings, with a direct line-in from a Denon tape deck to my Zoostorm PC. The source tapes are speech only, and so far the results are okay but disappointing. Whether I export to a wav or a FLAC file, in 32 or 16, the speech is noticeably lower in definition than the source. To the extent where it’s hard to understand even when cranked up. Doesn’t sound so much like over-loud distortion, as such, although there is a bit of clipping from time to time in these sample recordings. It just really sounds like a lo-fi digital version of the original, like all the fidelity couldn’t all be captured.

So I’m thinking my Realtek High Definition Audio onboard soundcard isn’t quite up to the job? I have a Behringer UCA202 U-Control Ultra low-latency 2 In/2 Out USB/Audio Interface on the way, so how do you rate its chances of producing better results? Is there anything else I can try/tweak in the meantime that might improve results?

The source tapes were mostly recorded on an analogue deck with two speakers - does that mean I should set Audacity to record in stereo, or…?

Ah, the joys of a cassette-converting newbie, eh? :slight_smile: Thanks in advance to anyone who can help!

with a direct line-in from a Denon tape deck to my Zoostorm PC.

If that’s a desktop/tower PC, a connection from the RCA line-outputs to the 2.5mm line-in on soundcard should be very good. (The mic input is “wrong”, and most laptops only mic-in.)

In general, the most important thing is to avoid clipping. An analog-to-digital converter “hard clips”, unlike analog tape which soft clips, so you should try to NEVER go “into the red” (although a small amount of hard-clipping may not be a problem). And unlike analog tape, there’s no need for a “hot” signal to overcome tape noise, so it’s OK to record at a much lower level. You can increase the level in Audacity after recording.

Whether I export to a wav or a FLAC file, in 32 or 16, the speech is noticeably lower in definition than the source.

How does it sound when you play back in Audacity before exporting? Any of those formats are fine… CDs are 16-bit/44.1kHz… That’s good enough for anything and far-far better than cassette and certainly good enough for spoken voice.

I have a Behringer UCA202 U-Control Ultra low-latency 2 In/2 Out USB/Audio Interface on the way, so how do you rate its chances of producing better results?

If you’ve got a desktop/tower with line-in, probably not. Usually if there is a “weakness”, it’s noise. If noise isn’t a problem (and the soundcard noise is probably less than the cassette noise) then a better interface probably won’t improve sound quality. Distortion and frequency response are usually better than human hearing.

The source tapes were mostly recorded on an analogue deck with two speakers - does that mean I should set Audacity to record in stereo, or…?

Since it’s voice only, it’s probably mono… Either the same sound on both sides or sound on only one side. It may be better to mix-down to mono but you may not hear any difference. A mono WAV file will be half the size. With MP3 it won’t make much difference because MP3 “knows” left & right are similar or identical, and FLAC does something similar.

Thanks very much for this, Doug, very helpful indeed. I am indeed using a desktop PC tower.

Just so I know, what is the reason for avoiding clipping like the plague? Is it as simple as just not wanting the loud parts to deafen you and get distorted, or is there another reason? Clipping doesn’t affect the overall recording, for instance, does it?

The actual Audacity project file sounded as disappointing as the exported wavs. But I’ve made a breakthrough! I’ve found that my results are 10 times better when I use the Denon’s headphones output, rather than the line-out. Now, the recordings seem to be pretty much the same as the tape! I was starting to think I was expecting too much from the tape-to-digital process, but now I’m really relieved. So I wonder why the output made such a difference? Do you reckon it just varies from deck to deck, in an idiosyncratic way? And thanks, yes, mono sounds just as good as stereo - the same, actually, which makes sense.

Denon’s headphones output

You should be careful with that. The headphone connection usually comes with a volume control somewhere on the unit. Once you get a transfer you like, don’t touch that knob, or mark it carefully so you can always go back.

Koz

Ah! Thanks, that’s scary. :slight_smile: Okay… so I’m seeing an Input Level twirler (probably not the right technical term), which is handily exactly half way between min and max. Also a Bias Fine twirler. Can’t see anything else that looks adjustable in terms of volume!

I also notice that I had Dolby B on, even though I can’t be sure the tape employed it. (This was a phone conversation recorded in the mid 90s, hooked up to an old tape deck, so I’m thinking I might have activated Dolby if I had the option, which I probably did…)

Here’s another question: I’ve been saving each project with 16 and 32 sample rates (at least I think that’s the right term), in case I ever want to burn them onto CD. But am I right in thinking that’s a needless use of disk space, because I will always be able to switch a 32 recording to 16 in future?

Thanks, Audacity creators, for this program. It’s so wonderful to be able to immortalise all my old tapes.

Audibly there is very little difference between 16-bit and 32-bit float. You can think of it as 16-bit format being “very nearly perfect” and 32-bit format being “very very nearly perfect”.
The main benefits of 32-bit float are mostly relevant when processing audio (applying effects). The number one benefit is that 32-bit float is not irreversibly damaged if you accidentally go over 0 dB due to processing.

Provided that your 16-bit files are not clipped (provided that the peak level is below 0 dB), and provided that you do not intend to apply further effect processing, then 16-bit is good enough and 32-bit float offers no advantages.

If you think that you may want to apply additional effect in the future, then technically there is a slight advantage to retain 32-bit float backups of the original material. In practice, the benefit is likely to be insignificant.

32-bit float files require twice as much disk space as 16-bit.
There is a compromise, which is to use 24-bit, which is higher resolution that 16-bit, 50% bigger file size, but like 16-bit it does not support over 0 dB.

One of the main benefits of 16-bit is that it is almost universally supported by software and modern hardware.

In short, if you need to have a perfect copy of the audio from an Audacity project, and disk space is not an issue, then use 32-bit float format. Otherwise, use 16-bit when exporting your completed projects.

Brilliant, thanks Steve! I’m still not clear on why clipping is an issue: does clipping affect an overall file in any meaningful way, beyond those individual moments being distorted?

I’m not sure what you mean by “meaningful way”.
Clipping is the distortion that occurs when peaks in the waveform go too high (or too low) and are “clipped” off. The lost peaks cannot be recovered (the “Clip Fix” effect can attempt to replace the clipped peaks by synthesizing the peaks, but that is usually far from perfect).

Thanks Steve. What I’m trying to clarify is, whether clipping negatively affects the file as a whole, or whether the negative effect is solely restricted to those lost moments of audio?

The damage is restricted to the clipped region.

I see, thank you sir. I seem to have got the hang of restricting the top level to -6db during recording, then using Normalize to take it up to -1, so all good!