Are modern recordings so bad that they would sound the same if recorded on a
cassette?

I've heard this said a number of times, and with my critical thinker running
in overdrive as it usually is, I just don't want to believe it.

I've seen this stated a number of ways, sometimes even suggesting that an
experienced audio professional wouldn't be able to hear the difference in a
close blind test.

I suspect that that reasonably good digital now is so pervasive, people have
forgotten how bad, even really pretty good cassette recordings are by modern
standards.

To shed some light on this I fired up a fairly modern and lightly-used
higher-end Sony consumer cassette recorder that I use quite a bit for
cassette transcription work. It generally sounds good to me. I use it at
least weekly, but this is light use compared to how a cassette machine would
be used by a consumer who primarily listened to cassettes. Its a
multi-function box, but the cassette part was rarely used.

I used a fresh Maxell XL-II tape. I recorded with Dolby B NR, as this was
more-or-less the defacto standard for cassette recordings. Since the deck is
auto-reverse, I ran some tests with the tape running in both directions. Of
course it measured a bit different in either direction, but on balance
neither was too bad. The channels were well-balanced, within a dB or better.

I set recording levels by recording 300 Hz tones at various levels in 3 dB
increments and then playing them back. The analysis was done in the digital
domain. With the tape stopped all electronic noise in the cassette machine
and test equipment was at least 10 dB below the noise level from a blank
fresh tape or an erased blank tape. So, the electronics weren't that bad.

To test the cassette's cleanliness at various levels, I filtered the tape
with a brick wall filter at 200 Hz to get rid of most hum and noise and
infrasonic junk. I first-order low-passed the at 20 KHz to set a noise
bandwidth. I notched the 300 Hz signal out in one channel, and bandpassed
the 300 Hz signal in the other channel The filter ran from 270 to 33 Hz
with -100 dB brick walls. I then examined the 300 Hz test tones and picked
out the two levels where the signal was 30 and 40 dB above the noise. IOW
about 3% THD+N and 1% THD+N. These levels were about 13 dB apart. I then did
the remaining tests with test signals (tones and music) recorded both of
these two levels.

I then recorded and played-back the PCABX test suite which is on the
current RAP CD set. It is composed of test tones and musical sounds. Once
back in the digital domain I level-matched the 1 KHz tones back to their
original levels along with the other signals in each test suite. The tones
were stable enough so that by averaging over about 5 seconds, I could match
levels within about 0.01 dB.

I listened to some preliminary results and immediately discarded all of the
tests done with peak levels set at the 1% THD+N point because the background
nose was so high that reliable detection took a less than trivial effort.
It was too easy. The SNR was about 40 dB.

The tests recorded at a level that corresponded to about 3% THD+N were not
so obviously detectable. They had a SNR that was in the 55 dB range which
seems to be about right if my recollections of how a good cassette machine
worked in the day of, are correct.

I then picked out the first PCABX musical sample called "castanets", edited
it for millisecond-level timing accuracy, and did a PCABX test. I obtained
16/16 reliable detection without any trouble at all. Remember, I'm 58 so the
high end response of my ears is not exactly pristine. This was a slam dunk.
Cassette record/playback is really quite clearly audible.

At any rate the PCABX samples you can download from www.pcabx.com are
themselves pretty pristine. So these tests don't really relate to the
original claim.

Can anybody name a modern CD that is so trashed that I would find hard to
detect if I bounced it through this typical reasonably high quality cassette
machine in reasonably good shape? How about Californification or some
Smashmouth? Madonna?

BTW, if there's any question in anybody's mind about how much cassette
record/playback trashes music compared to good MP3 processing at 128 kb or
above, there's no comparison. I actually have to sometimes actually work at
it to hear MP3 processing, while this cassette-based listening test was no
work at all. But, its a different kind of trashing.