HydrogenAudio

Hey HA experts, I've purchased a mediocre "Walkman" which outputs to USB (https://www.amazon.it/gp/product/B078W4Y6NS/) and therefore allows me to record/rip on a computer, I just need to transfer some demos, rehearsals and bootlegs of my 80s band (which went nowhere course) all abysmal quality in the first place, I'm stuck with Windows 10 for this little project and I'm using Audacity, care to help me get it right and make some decent files please?

Here are the first two questions:

• The device is recognized as a USB mic like this:
  (https://i.imgur.com/U4XtVDQ.png)
  Should I match the device's default format with Audacity's sampling?
  (https://i.imgur.com/WzHrEix.png)
• Should I lower the volume of the source tape until no clipping is detected by Audacity's Recoding Meter Monitor and/or is there a reference level I could use?

Thank a bunch.

Does this tape machine provide specs on its output format? Its default output is probably its best output. The recording program should be set to the same sample rate it is receiving. Ideally the bit depth would be floating point, for the greatest ease of control for cleanup, editing, and mixing. If you don’t intend to do any of that, then 16 bit is adequate.

The output level control, should be set to avoid clipping although if there is only infrequent clipping that is not too extreme, you probably will not be able to notice it.

It's this gizmo: https://www.amazon.it/gp/product/B078W4Y6NS/, no specs of the output format mentioned anywhere, is there a way to determine them please?

Should I lower the volume of the source tape until no clipping is detected by Audacity's Recoding Meter Monitor and/or is there a reference level I could use?

It's the analog-to-digital converter (built-into the cassette player) that will clip (distort) if the analog level is too high,    In these devices you can't control the analog level into the ADC so you just have to HOPE it's calibrated properly and leaving plenty of headroom.     You can boost the level after recording but of course lowering the level doesn't remove distortion.

Note that lowering the volume will "hide" the clipping from Audacity...   Audacity is just checking the levels,   It's not looking at the wave shape. 

Looks like they have two near-identical models. A search engine gives me claim of mp3 and claim of 44.1/16 or 48/16.

But the AudioLava software is a declicker/de-noiser, right?

AudioLava also appears to also be a recording application.
https://download.cnet.com/AudioLava/3000-2170_4-10637884.html

The Amazon ad linked to seems to indicate mp3 output from the device. This is not an editable format. If that is all you can get, it would have to be decoded before any program could work on it (except, of course, for the limited operations of programs like mp3DirectCut).

There are various youtube videos on these types of devices. I don't know if this particular one is covered but there may be useful information in some of the how-to presentations.
https://www.youtube.com/watch?v=Ie6dyZ0P4Rw
https://www.youtube.com/watch?v=lOn-vU63A5Y

 If the device only outputs undecoded mp3, recording it as .wav would not produce anything useful. I would capture the mp3 then decode it to wav KEEPING the original file, at least until EVERYTHING you ever might want to do with it is accomplished.

Waitaminute;

I have to apologize for not catching the point here:

The device is recognized as a USB mic like this:
(https://i.imgur.com/U4XtVDQ.png)

So it seems it is simply a USB microphone yes.  I've never heard of any of those doing stuff like MP3 conversion (I don't think that is a thing in USB-audio?!).

Rather their ADC apparently output straightforward PCM to the USB port - and and might support several sampling rates. In this picture, is the "2 channel, 24 bit, 48000 Hz (Studio Quality)" greyed out, or can you get other options by pulling down? If not, then 48k it is.

However, I see people have run into trouble: https://stackoverflow.com/questions/50396224/how-to-get-audio-formats-supported-by-physical-device-winapi-windows

So it seems it is simply a USB microphone yes.  I've never heard of any of those doing stuff like MP3 conversion (I don't think that is a thing in USB-audio?!).

Rather their ADC apparently output straightforward PCM to the USB port - and and might support several sampling rates. In this picture, is the "2 channel, 24 bit, 48000 Hz (Studio Quality)" greyed out, or can you get other options by pulling down? If not, then 48k it is.

However, I see people have run into trouble: https://stackoverflow.com/questions/50396224/how-to-get-audio-formats-supported-by-physical-device-winapi-windows

These things record mp3 into USB thumb drives.
https://www.amazon.com/Rybozen-Cassette-Portable-Converter-Recorder/dp/B083VTXT2G

For the Audacity settings, just record in 16-bit, and convert to 32-bit after recording for editing, to reduce initial file size and minimize disk wearing.

These things record mp3 into USB thumb drives.
https://www.amazon.com/Rybozen-Cassette-Portable-Converter-Recorder/dp/B083VTXT2G

That is a slightly different one, it seems.

Quote from: bennetng on 2022-01-03 02:26:21

These things record mp3 into USB thumb drives.
https://www.amazon.com/Rybozen-Cassette-Portable-Converter-Recorder/dp/B083VTXT2G

That is a slightly different one, it seems.

That's why I said these things, not this thing. I don't think OP's model can directly output to mp3. The English version of Amazon says "Convert Your Tapes to MP3 only 2 steps" as well.
https://www.amazon.com/gp/product/B078W4Y6NS/

If the cassette player is driverless and has no thumb drive connectivity, the only way I can think of to output mp3 to a Windows PC directly is to have built-in storage and expose itself as a disk drive to Windows, like what smartphones do.

It has headphone output. The best thing to do imo would be to connect the walkman's headphone output into your computer's line-in input.
(Do not use the microphone input, you will fry it, ask me how I know lol...)

That way you will bypass the walkman's built-in ADC (probably shitty ADC) and you can record in pure PCM without any compression.

Make sure to reduce the volume to minimum and start increasing it before it starts clipping.
The higher the level, the less noise you will have, but always leave couple of dB of headroom, do not push it.

Be careful, some mixed cassette tapes have music recorded at different level, one song can be quiet, other song can be very loud.
This is not "set and forget" process. You will need to monitor whole recording session.

Once you have finished recording, you will hear that your tapes sound like shit.
Now you will need to play with noise reduction and I highly recommend to use some high frequency enhancers to bring
back your music to life.

If your channels are not balanced (one channel is louder) you can simply amplify it to make it sound equally loud.

If you have cassettes recorded with Dolby NR, I can recommend this program. I had very good results with it.
https://sourceforge.net/projects/dolbybcsoftwaredecode/

I can see that you are doing this as a hobby so do not expect the "CD quality" sound.
Cassette tapes are very hard to work with. I don't even know where to start. From different types, materials, azimuth adjustment and so on...

Note that lowering the volume will "hide" the clipping from Audacity...   Audacity is just checking the levels,   It's not looking at the wave shape.

Then should I use the max volume possible or not please?

Besides Audacity's Recoding Meter Monitor detecting clipping the wave is taking up all the available area, it becomes a full blue histogram at high volume.

But the AudioLava software is a declicker/de-noiser, right?

Just a piss-poor audio converter, I'm not using it.

Rather their ADC apparently output straightforward PCM to the USB port - and and might support several sampling rates. In this picture, is the "2 channel, 24 bit, 48000 Hz (Studio Quality)" greyed out, or can you get other options by pulling down? If not, then 48k it is.

(https://i.ibb.co/c3ntrNr/3.png)

Quote from: bennetng on 2022-01-03 02:26:21

These things record mp3 into USB thumb drives.
https://www.amazon.com/Rybozen-Cassette-Portable-Converter-Recorder/dp/B083VTXT2G

That is a slightly different one, it seems.

Yep, not my thingie.

It has headphone output. The best thing to do imo would be to connect the walkman's headphone output into your computer's line-in input.
That way you will bypass the walkman's built-in ADC (probably shitty ADC) and you can record in pure PCM without any compression.

I don't have a line-in input and it's very low quality source material, I'm fine with its crappy ADC for this little project, thanks.

Make sure to reduce the volume to minimum and start increasing it before it starts clipping.
The higher the level, the less noise you will have, but always leave couple of dB of headroom, do not push it.

Sounds about right to me, OK.

Be careful, some mixed cassette tapes have music recorded at different level, one song can be quiet, other song can be very loud.
This is not "set and forget" process. You will need to monitor whole recording session.

It's two demos, one rehearsal and two live bootlegs.

Once you have finished recording, you will hear that your tapes sound like shit.
Now you will need to play with noise reduction and I highly recommend to use some high frequency enhancers to bring
back your music to life.

Anything built-in in Audacity I can use or maybe a free plug-in please?

If your channels are not balanced (one channel is louder) you can simply amplify it to make it sound equally loud.

Noted, thanks.

If you have cassettes recorded with Dolby NR, I can recommend this program. I had very good results with it.
https://sourceforge.net/projects/dolbybcsoftwaredecode/

I'm bookmarking this for eventual future use, no Dolby NR here at all on these tapes.

I can see that you are doing this as a hobby so do not expect the "CD quality" sound.
Cassette tapes are very hard to work with. I don't even know where to start. From different types, materials, azimuth adjustment and so on...

I'm turning 50 in a few months and I wanna use my 80s band tunes as the soundtrack to the party, quality as already underlined is very very poor in the first place.

Thank you very much for taking the time to help me out, appreciated.

Just a piss-poor audio converter, I'm not using it.

That is not true according to the software itself. It records, it does various kinds of audio transforms for improving the finished product. Those tranforms (noise reduction, declicking, convolutions, etc. probably are of lesser quality than some other programs.
However, for recording uncompressed audio, all programs that are not defective produce exactly the same result (for the bit depths and sample rates they handle). They simply accept samples from the ADC and writing them to disk (generally displaying information about the process as it occurs).

I don't know about the  tape player you have but many  similar tape players clearly contain ADCs and mp3 encoders. There is no other way they could output mp3 to a flash drive as their ads claim (no computer necessary, etc). The required electronics are easily small enough to fit inside. for those that can be used as digital input to a recording program on disk. Their analogue output (via the headphone jack) may be of lesser quality than their ADC output.

While the device probably does not have dolbyB, if it does you should test record some with it turned on vs turned off, even if the original recording did not used dolby. You may be happier with the noise reduction it provides when on.

A bit of test recording, with varying parameters) before you get down to the real project, is always a good idea.

That is not true according to the software itself. It records, it does various kinds of audio transforms for improving the finished product. Those tranforms (noise reduction, declicking, convolutions, etc. probably are of lesser quality than some other programs.
However, for recording uncompressed audio, all programs that are not defective produce exactly the same result (for the bit depths and sample rates they handle). They simply accept samples from the ADC and writing them to disk (generally displaying information about the process as it occurs).

I'm gonna stick to Audacity.

I don't know about the  tape player you have but many  similar tape players clearly contain ADCs and mp3 encoders. There is no other way they could output mp3 to a flash drive as their ads claim (no computer necessary, etc). The required electronics are easily small enough to fit inside. for those that can be used as digital input to a recording program on disk.

This is the one I have https://www.amazon.it/gp/product/B078W4Y6NS/ and conversion to MP3 is done by the software.

Their analogue output (via the headphone jack) may be of lesser quality than their ADC output.

Interesrting.

While the device probably does not have dolbyB, if it does you should test record some with it turned on vs turned off, even if the original recording did not used dolby. You may be happier with the noise reduction it provides when on.

No Dobly NR available.

A bit of test recording, with varying parameters) before you get down to the real project, is always a good idea.

For sure.

Getting the raw audio is always an advantage. It gives you the greatest leeway in processing whichever way you want.

Any way to tell which is the default ADC output?
The one I see in the recording device properties hence stereo 24-bit 48000 Hz?

Should I acquire at that format or use 32-bit float in case I wanna run some cleanup filters/plugins please?

Short answer:
Configure Windows for 24-bit / 48 kHz.
Record as 32-bit float in Audacity because you will be editing your files, and all editing is done in 32-bit float internally.
Even if you record at lower bit depth (to save disk space) you will again need to convert it to 32-bit float to do some editing so you actually won't save any space at all.
Make sure to change sampling rate in Audacity in the bottom left corner to 48 kHz before you start recording.

Kinda off-topic:
Sampling rate is a little bit complicated. I do remember having a device that supported lots of different sampling rates
(32, 44, 48, 88, and 96 kHz) but all of them except the 48 produced noise.
Turned out that card worked at 48 kHz natively and all other sampling rates were actually running on (bad) internal resampler.
The reason why I recommended to record using the line-in input is because I have no idea what kind of ADC your walkman has.
Have you ever bought Chinese 64 GB flash drives and then you found out later that real capacity is actually 8 GB?
Maybe the ADC chip is actually 8-bit and it works at 24 or 32 kHz and it was simply masked to show higher specifications?
By using line-in you will bypass all of that except the amplifier stage. Most amplifiers start "cracking" at very loud levels,
but we are not going to use levels that high so there should be no problems.

Don't be afraid by what I posted above. Everything is going to be fine. I just wanted to share my thoughts =D

Any way to tell which is the default ADC output?

That's generally hidden and sometimes the published specs aren't true...   The Windows drivers will re-sample so you can record a 24-bits/192kHz file with a with a cheap soundcard that's really working at 16-bit/44.1kHz.   Similarly, you can play a "high resolution" file on any-old cheap soundcard and Windows never tells you what's going-on... 

The actual hardware is 16 or 24-bits.   By default, Audacity works "internally" at 32-bit-floating point.   There are technical advantages to processing in floating-point.

And with floating point there is essentially no upper (or lower) limit so you can do something like boost the bass to the point where Audacity shows red for (potential) clipping but the waveform isn't actually clipped (yet) so you can simply lower the volume as necessary before exporting.    (You will also clip your DAC if you have a file/data that goes over 0dB and you play it at "full digital volume".)

When you export finished the file, "CD quality" (44.1kHz, 16-bit) is generally better than human hearing so it's good enough for almost anything, and certainly good-enough for analog cassette (or vinyl).     You can use 48kHz if it makes you "feel better" and with uncompressed WAV you'll just get a proportionally-larger file.   And a 24-bit file is 50% larger than a 16-bit file, etc.

If you want to make MP3s it's probably a good idea to "archive" a WAV file.   And if you're doing any noise reduction or EQ, etc., you might want to save an original untouched WAV just in case you want to go-back and start-over.   (You probably don't need to save an Audacity Project.)

@Markuza97 I don't have a line-in input on that Windows machine.
@DVDdoug I'm going to export CD quality (maybe 48kHz) and compress with WavPack.

Crunch time: USB thingie in Windows set @ 24-bit 48kHz, Audacity with sample rate 48kHz and sample format 32-bit float (it's the default, nothing to configure, right?), OK?

Just as with social media, Windows is now more about you being the product rather than giving you the full power of a computer. It used to be that professional and semi-professional soundcards had drivers that bypassed all the Windows manipulations, letting a recording program receive exactly what the ADC output. I think in Windows7 MS started making that impossible in order to protect their real clients' interests.

I haven't paid attention for some time; I don't know what professionals use now. Possibly recording is mostly done on other systems or stand alone recorders. Once you have the file, I suppose Windows will still do what it is told.

It also used to be that MB ADCs and DACs were mostly poor quality but apparently the technology got so cheap that many board manufactures started using better chips. I still have a couple of older computers using older OS versions with quality soundcards plugged into add-on slots, but I don't know if such cards are even produced these days since MS made it more or less impossible to fully utilize them.

Audacity with...32-bit float[/b] (it's the default, nothing to configure, right?)

Right, but if you want to confirm:

Edit > Preferences > Quality > Sampling

A true ASIO driver provided by the designer of the sound card should bypass Windows components (not an emulated driver on top of Windows). I observed ASIO to be much more reliable compared to the Windows Server 2008 sound system on an underpowered CPU. Windows invented a new priority management system (MMCS) to make sound work decently.

It's possible that the cheap device would distort at its maximum level. With a good sound card, it is not critical to reach full scale level. The tape noise will dominate anyway. The sampling rates of both devices will not be synchronized, so it is pointless to try to match them.

I would avoid noise reduction. It will likely result in muffled tone or bubbling FFT artifacts. Maybe reduce the noise slightly by 12 db or so, and fade the begnning and end mathcing the speed of the music.

NR can be done quite successfully without any artifacts but good NR facilities have a number of important parameters that must be selected well to avoid making thing worse rather than better. It also can sometimes be difficult to do very much NR without effecting the "atmosphere" of the recording space or 'effects' of the older equipment used, particularly for older on site recordings where the building was an important aspect. I probably experimented a couple hundred hours to learn how my software gave the most satisfactory results.

Some fact checks about bit-depth and file size.

Record 20 seconds in 16-bit, trimmed to 10 seconds, then peak normalize:


Same but 32-bit float:


While software these days often operate at 32/64 bits, the meaning of setting the project bit-depth is for the temp files. Trimming and peak normalization don't really require high precision. For me, I always keep the original 16-bit non-normalized and non-trimmed files, and examine the COPIES of the original files with different restoration software and plugins, and of course, at higher bit-depth. Cassettes are 10 bits at best so there are 6 bits for self-dither and I see no reason to record at higher bit-depth, despite how cheap storage devices are.

The more advanced DAWs are often non-destructive so they don't really create any temp audio file until you perform a bounce/export action, the undo data are just parameters rather than actual audio data.

For me, the "restored" files are actually disposable. When another software/plugin shows up and I am happier with the newer results, I often delete the old ones.

Cassettes, much like vinyl, are still an analog medium, so you want to record as much precision as you can handle, to better facilitate any cleanup you have planned for the recording.

For those who want to try, here is a 24/96 raw recording of a TEAC cassette deck to RME Mulitface II, and obviously 32-bit float cannot be saved as flac.

If anything, for analog recording, the source equipment (tape player, turntable etc) is way more important than digital recording format. So if there is money to spend, better invest in these things rather than buying a bigger SSD.

Please don't hijack this and turn into something else: I have very bad quality source material recorded very badly on worn out tapes, I merely need to make a decent copy in digital format I can reproduce on computer, smartphone and DAP.

I've now acquired the audio of the first tape with the USB thingie in Windows set @ 24-bit 48kHz and Audacity @ 32-bit float 48kHz.

Next steps please:
1) A bit of clean up: any built-in filters or something for tape sources? Any plug-ins? Should I bother with NR even though none was used to record the source material?
2) Audio split into files or CUE sheet?
3) Export to WavPack @ 16-bit 44KHz directly or a different format to be able to edit losslessly the capture again?

Thanks.

Then it even makes less sense about bit depth, and it is not "hijacking", because YOU asked about bit depth in the first place.

1) For the noise reduction, Audacity has a built-in one, and can be found in the effect menu, but I suppose you can find more reference in the manual as well, it is more about tweaking with the parameter to fit your taste. There are commercial ones from iZotope as well, and can be used in standalone mode as well.

3) If you export the raw recording at 16-bit, it is essentially no different from recording at 16-bit in the first place, apart from the sample rate conversion from 48 to 44.1. Just to avoid being called "hijacking" again, sample rate conversion is not a mathematically lossless process. If you must resample the raw recording, export as 32-bit float will be more mathematically accurate, and WavPack supports 32-bit float as well.

You are correct. We hijacked the thread but I cannot really blame us. This is audiophile forum, we are always looking to squeeze as much quality as possible.

Based on what you said (low quality recording, worn out tapes, cheap walkman) I wouldn't even bother saving the lossless copies.
I think there is too much damage that even the best paid software on the market won't be able to repair.
I would just do it once and forget about it. Here's what I would personally do:

1. Split the recording (different songs) and export each song as 32-bit float / 48 kHz. (This is lossless process.)
2. Use the built-in noise reduction tool in Audacity like others mentioned. Find settings that sound best for each song.
Do not overuse noise reduction. If you do, you will get "synthetic" sound. Little bit of noise is perfectly fine.
3. I don't really like tag-based ReplayGain. Most software doesn't support it. I would personally use loudness normalization effect
in Audacity to get equal loudness. If you have channel imbalance like I mentioned in my previous post, you can normalize each channel separately.
4. Apply fade-in and fade-out for smooth transitions.
5. Resample to 44 kHz.
6. Export as MP3 using LAME (Audacity is using LAME internally) at V2 preset. Your files will play anywhere and they will have good size.

It all comes down to one thing: testing.
You will need to do some tests with different settings to find what sounds best for you.
Not a single person on this forum can give you definitive answer on what you should or shouldn't do.

Next steps please:
1) A bit of clean up: any built-in filters or something for tape sources? Any plug-ins? Should I bother with NR even though none was used to record the source material?
2) Audio split into files or CUE sheet?
3) Export to WavPack @ 16-bit 44KHz directly or a different format to be able to edit losslessly the capture again?

1) Audacity has a "Noise Reduction" effect that can be effective: https://youtu.be/ZCTUwi7jTao
2) It doesn't matter, "tracks" (files) are what's typically done.
3) Save the Audacity "Project" after exporting to whatever final format you want, e.g. MP3-320.  If you need to make future edits, just reopen the project and it returns to where you left off.

Next steps please:
1) A bit of clean up: any built-in filters or something for tape sources? Any plug-ins? Should I bother with NR...?

I'd try noise reduction, but there can be artifacts (side effects), especially when the noise is bad and "the cure can be worse than the disease" so listen carefully to the results and maybe play with the settings and decide for yourself if you like the results.

even though none was used to record the source material?

Audacity's (and Izotope's, etc.) noise reduction doesn't rely on encoding/decoding like Dolby NR. 

You may want to experiment with equalization (to adjust the "frequency balance").

You'll probably also want to normalize ("maximize" the levels).   Do this as the last step after all other editing/processing.  If this is an album and you want to maintain the relative loudness of loud & quiet songs, normalize the whole thing at once.   Otherwise you can normalize one song at a time.   (Normalization is simply a linear volume adjustment and it has no effect on sound quality.)

2) Audio split into files or CUE sheet?

Do you need a cue sheet?   I don't make a cue sheet unless I'm burning a CD.   With Audacity, I just select/highlight one song at a time and then Export Selected Audio, one song at a time.  (I usually select a little extra at the beginning & end, then re-import and trim.)   There is also a way of labeling the tracks and then using Export Multiple to make multiple files at once.   (The details are in the Audacity manual.)

Audacity can "tag" the files (artist/album/title, etc.) but if you want album artwork you'll need a different tool.  I use MP3Tag (which also works with other formats, not just MP3).

3) Export to WavPack @ 16-bit 44KHz directly or a different format to be able to edit losslessly the capture again?

Yes.   It's always a good idea to keep a lossless archive copy, especially if you are making MP3s or other lossy files.   

When you open a compressed file in Audacity (or any "normal" audio editor) it gets decompressed.   If you re-export as MP3 you are going through another generation of lossy compression and some damage does accumulate.  (AAC is immune to multiple-generation damage, but you might want a different format sometime in the future.)

Based on what you said (low quality recording, worn out tapes, cheap walkman) I wouldn't even bother saving the lossless copies.

Just in case I get a bit familiar with Audacity and find better ways to clean up the audio later on I may then wanna use on the files already exported, no?

1. Split the recording (different songs) and export each song as 32-bit float / 48 kHz. (This is lossless process.)

Export to WAV?
If yes then I'm compressing them with WavPack which I already use for my CD rips as it supports 32-bit float.

3. I don't really like tag-based ReplayGain. Most software doesn't support it. I would personally use loudness normalization effect in Audacity to get equal loudness.

Isn't normalization a lossy process please?

5. Resample to 44 kHz.

Do you mean in Audacity before exporting or via exporting please?

6. Export as MP3 using LAME (Audacity is using LAME internally) at V2 preset. Your files will play anywhere and they will have good size.

I'll stick with WavPack hybrid (store correction files on NAS and use lossy files), thanks.

You will need to do some tests with different settings to find what sounds best for you.

Will play with it a bit for sure.

Thanks a lot for your help.

1) Audacity has a "Noise Reduction" effect that can be effective: https://youtu.be/ZCTUwi7jTao

Thank you very much for the link to a quick tutorial!

3) Save the Audacity "Project" after exporting to whatever final format you want, e.g. MP3-320.  If you need to make future edits, just reopen the project and it returns to where you left off.

So you're suggesting saving the .aup3 files cause exporting to WAV or another lossless format isn't the same thing?

I'd try noise reduction, but there can be artifacts (side effects), especially when the noise is bad and "the cure can be worse than the disease" so listen carefully to the results and maybe play with the settings and decide for yourself if you like the results.

Will do.

You'll probably also want to normalize ("maximize" the levels).   Do this as the last step after all other editing/processing.

But it's a lossy process, no?

There is also a way of labeling the tracks and then using Export Multiple to make multiple files at once.   (The details are in the Audacity manual.)

That sounds very interesting, will look it up.

Audacity can "tag" the files (artist/album/title, etc.) but if you want album artwork you'll need a different tool.  I use MP3Tag (which also works with other formats, not just MP3).

I'm gonna tag the WavPack files only with its dedicated tool WvTag, thanks.

It's always a good idea to keep a lossless archive copy, especially if you are making MP3s or other lossy files.

My questions is the same one I've just asked to Apesbrain: do I need to keep the Audacity project in .aup3 format or is it the same to export to a lossless format @ 48kHz 32-bit float? I mean, does the Audacity .aup3 file contain anything the lossless export won't please?

So you're suggesting saving the .aup3 files cause exporting to WAV or another lossless format isn't the same thing?

I find it convenient to save my Audacity Project (AUP3) regardless of the export format.  That way, if I ever want to come back and change something about my files, I can go back to the original Audacity 32-bit "workprint".  For me, it's just easier than what you've described.  Quality-wise, there is no difference.

But it's a lossy process, no?

No, it's not lossy. 

If you export to an integer format after normalizing (or any other processing) there is some tiny-tiny-rounding but it's not considered to be a lossy process and it doesn't audibly degrade the sound.   Analog amplification isn't "perfect" either.   ;)

As you already know, the original cassette tape is your weak link...

do I need to keep the Audacity project in .aup3 format or is it the same to export to a lossless format @ 48kHz 32-bit float?

Personally, I'm usually doing "simple things" similar to what you're doing and I rarely make an Audacity project.  You can keep an AUP3 project file but I'd also recommend a WAV or FLAC.   People on the Audacity forum are occasionally getting corrupted AUP3 files (there were also occasional issues with the old AUP format).    AUP3 files also can't be saved on FAT or FAT32 drives (i.e. thumb drives) and some people have trouble with cloud storage. 

WAV is simpler and more "foolproof" and it's universal so you should have no problem opening or playing a WAV file with whatever software you have 20 years (or more) from now.   FLACK will also probably still be around.    FLAC doesn't support floating-point but IMO you don't really need floating point for archival storage.

Normalization, except "bit-shift" (doubling), is not mathematically lossless. It is rather verbose to add "mathematically", but some may consider "percetually lossless" as "lossless" as well. Doubling is to increase about 6.0206dB of volume instead of some arbitrary values, it may either cause clipping or over, so not a very useful action anyway.

I have a feeling that OP while agrees that the source and equipment are of low quality, but still prefer "mathematically lossless". So here are some important points. Your walkman and even hi-end audio ADCs don't record in floating point, so you are in fact getting 24-bit data into Audacity, assumes:
[1] the Walkman is really 24-bit capable,
[2] you have not performed any normalization
[3] you have dithering disabled in Audacity's perferences
Then even if the project is 32-bit float, you can still export 24-bit files mathematically lossless.

Of course, if you prefer to use 32-bit float WavPack, it is perfectly fine too.