top of page

Technical Challenges Posed by Personal Recordings  (October 2019, updated February 2020)

 

I enjoy converting commercial records and cassettes.  The conversion, noise reduction, track splitting, titling, and metadata processes are all straightforward, and I’m exposed to music that I’d otherwise be unlikely to encounter.  But I really enjoy working with personal recordings for three reasons:  customers are emotionally invested in the audio content; they’re willing to work with me in determining where tracks should be split and what they should be called; and I’m often surprised with new challenges. 

 

Personal recordings are mostly cassettes or reel-to-reel tapes, but I occasionally encounter nitrocellulose lacquer records, microcassettes, and minicassettes.  In what follows, I’ll present several narratives about challenging projects and address the issues I encountered, why I think they occurred, and what I did about them.  You may recognize some of these stories from earlier articles, but others are new.  Most involved tape speed or noise.

 

Tape Speed

 

Different Recording Speeds

A customer’s father had made numerous recordings using a reel-to-reel tape deck.  Most decks intended for home use offered two tape speeds, 7.5 ips (inches per second) and 3.75 ips.  The faster speed yields recordings with greater fidelity while the slower speed would double the recording time.  He used both speeds, typically the faster to capture family members playing music and the slower to capture conversations.  In converting the tapes to digital, this is easily accommodated by selecting the playback speed that matches the recording.  When different speeds were used on the same tape, I could stop the playback, backstep the tape, change the speed, and continue with the conversion.

 

He also used a speed of 1.875 ips, a speed available only on some older decks, to capture a long musical performance, sacrificing fidelity in the process.  Neither of my tape decks offers that speed, so I captured the recording at 3.75 ips and used software to reduce the audio speed by 50%, thereby lowering the pitch by one octave and correcting the tempo.  A similar approach can be taken should I encounter recordings at 15/16, 15, or 30 ips—speeds not normally found on consumer tape decks.

 

Speed Variability Between Devices

A friend had recorded conversations with elderly relatives on microcassettes and asked me to digitize the recordings and clean up the noise.  He loaned me three of his players with the intent that I could use the one that produced the cleanest sound.  The recordings reflected what I would consider normal voice pitch and tempo on only one of the three.  The voices were either too high and fast or too low and slow on the other two devices, regardless of whether the fast or slow tape speed was selected.  This suggested to me that there is variance in the manufacturing process, and that microcassette recordings will sound acceptable as long as they’re played on the same device that recorded them.  I suspect this variance may also occur with minicassettes and with inexpensive open-reel tape recorders.  Fortunately, all my friend’s recordings appear to have been made on the same recorder, and that’s the one I used to play the tapes during conversion.

 

Weak Battery

A customer requested conversion of a cassette featuring recollections by his grandmother.  The conversion and noise reduction were unremarkable   After listening to the CD, however, he indicated that his grandmother “sounded like a chipmunk” in the second track (corresponding to the cassette’s second side).  During our ensuing conversation, it came to light that the original recording was made on one machine, and that he had used a portable device to record his cassette while the original tape was played on the original device.  The explanation of the “chipmunk” effect of increased pitch and tempo that best fit the situation was that the battery in his recorder had weakened when the second side was being recorded.  A weak battery would reduce the tape transport speed, and thus my cassette deck played that side faster than the speed at which it was recorded.  I offered him samples that were slowed down by 5%, 10%, 15% and 20%, and he picked the one that best resembled how he remembers his grandmother’s voice.

 

Tape Transport Mechanism

One of biggest challenges I’ve faced involved a collection of 3” open-reel tapes.  Some of them played fine on my tape deck, and conversion proceeded normally.  With the others, however, the speech got lower, and the tempo slower, as playing progressed.  At first I thought there might be a problem with my deck:  Increased drag on the feed reel due to the reel hub’s small diameter might be causing the tape to slip.  The problem replicated on a second deck, as well as when I transferred the tape to a larger reel.  I felt relieved that it wasn’t my deck but still perplexed as to what might be causing the problem.

 

At the time, I was reading about wire recorders, which used the take-up spool to transport the wire.  In contrast, reel-to-reel and cassette recorders generally use a capstan (a cylinder revolving at a constant speed) and pinch roller (which holds the tape against the capstan) to move the tape across the heads.  It occurred to me that the recorder used to create these cassettes, an inexpensive device manufactured in the mid-1960s, might be using the take-up reel to transport the tape instead of a capstan/pinch roller mechanism.  Whether this hypothesis was correct remains undetermined, but it was consistent with what I was hearing from the converted audio.  As more and more tape was wound around the take-up reel, the diameter, and consequently the circumference, of the accumulated tape grew.  This increased the tape’s speed across the record head, which in turn progressively slowed down the audio when the tape was played at a constant speed.

 

There may be software available that can address this issue of progressive change in tape speed, but none was available to me at the time, and I’m not convinced that it would have satisfactorily addressed the problem.  I proceeded by listening to the audio to determine where the tempo and pitch changes were first detectable, splitting everything before that point into a separate file, and increasing the speed of the remainder of the audio.  I did this iteratively and wound up with up to a dozen segments for each side of each tape.  To avoid an obvious change in tempo and pitch from one segment to the next, I would select a point where the person speaking paused or where a different person began speaking.  The segments averaged about 30 seconds in length; some were as short as 10 seconds, and a few lasted up to a minute where singing was involved.  Although the voices sounded more or less normal to me, I didn’t know any of the people speaking.  So, I ran all the segments past the customer (who was present when these recordings were made) and tweaked the speed of each segment in response to the customer’s feedback.  After the speed adjustments had been completed, the segments were joined back together to create the final audio track.  Though this was labor intensive, the end result was worth it to the customer.

Update (February 2020):  From time to time, I explore different audio editing packages to see how they stack up against those I use and if there’s anything new that they offer.  I came across a function in one such package that progressively changes an audio track’s playback speed and, concomitantly, pitch.  It’s a natural for addressing the situation I described above, and it’s now in my “arsenal.”

 

Noise

 

Permit me to digress a moment to present some background information.  Cassette tape is by nature a noisy medium because of inherent hiss, and special processes can be applied to minimize that noise during playback.  Dolby B and Dolby C are two noise reduction approaches that you may have encountered.  What noise remains, termed the “noise floor,” has such small amplitude that it’s virtually inaudible when playing a commercially produced cassette.  Another way to view this is the higher the signal-to-noise ratio (i.e., more of the recorded audio and less noise), the better the recording will sound.

 

Reused Tape on a Device Without an Erase Head

I’ve had numerous projects where customers provided me with recordings created on inexpensive devices using a tape that had something else previously recorded on it, and such projects can indeed be challenging.  Here’s why.

 

Better reel-to-reel and cassette recorders incorporate an erase head in addition to the record and playback heads, and its function is to remove magnetization on a tape before the tape passes over the record head.  On the other hand, inexpensive recorders, including those for minicassettes and microcassettes, do not have an erase head and instead rely on the new signal being recorded to overwrite what’s already on the tape.  Unfortunately, this is less efficient than employing an erase head.  The noise floor tends to increase in volume each time a tape is rerecorded, so much so that it is audible in all but the loudest parts of the new recording (and sometimes even then).  In extreme cases, one can hear content from an old recording during soft parts of the new recording, especially during periods of silence.

 

Regardless of whether the new recording is voice or musical instruments, removing the unwanted noise from the digitized audio is especially problematic.  The noise floor can be reduced, but only to a point.  Overaggressive noise reduction will introduce artifacts such as ringing tones or voices that sound “metallic.”  In situations like these, I often create three versions of a passage:  one that reduces the noise floor up to the point just below where artifacts first appear, one that aggressively reduces the noise floor but where artifacts are obvious, and something in between.  I then offer the customer a choice, accompanied by my recommendation and the reasoning behind it.

 

Low Signal-to-Noise Ratio

A closely related situation I’ve encountered occurs when the microphone is too far away from the speaker, the record level is set too low, a speaker’s head is oriented away from the mic, or a speaker’s voice is very soft compared to that of other voices being recorded.  In each of these, the problem is a low signal-to-noise ratio, where the voice almost disappears into the noise floor.  The same holds true for musical instruments, though this occurs less frequently and often in conjunction with a reused cassette.  My approach to dealing with this problem is the same as described just above.  Usually, my recommendation is to avoid any artifacts.  In one case, though, the tape noise was so pervasive that it was clearly audible throughout the recording, including its loud passages.  The customer agreed with removing the noise and putting up with a few seconds containing artifacts.

 

Record Level Too High

One would think that the signal-to-noise ratio could be increased by placing the mic directly in front of an instrument or the speaker (or the speaker with the softest voice) and cranking up the record level.  These work up to a point, but doing so risks introducing a separate problem:  tape saturation.  There’s an upper limit to how much the metal or metal oxide particles on a tape can be magnetized.  When approaching that limit, let alone reaching it, the strength of the magnetization that the particles can retain deviates from the strength of the signal.  The result is an altered sine wave—and distortion. Unfortunately, nothing can be done to remove that distortion, including lowering the playback volume.

 

Tape Transport Noise

While digitizing the microcassette tapes I mentioned earlier, I noticed a persistent whine throughout the recordings.  Since the recordings were made at different times and in different places, I could rule out environmental sources.  Its consistency suggested that the culprit was the recording device:  It was recording the noise from its own tape transport mechanism.  This issue can occur with any portable tape recorder with a built-in microphone—cassettes, microcassettes, minicassettes, and open-reel tapes.  The solution was to apply a “notch filter,” equalization that significantly reduced the volume of a very narrow frequency band centered on the frequency of the whine.  The resulting audio quality was minimally affected.

 

Other Challenges

 

Backwards Recordings

One of my projects involved several stereo cassettes, including one oddball—a cassette with separate material recorded in the left and right tracks of both Side A and Side B (yielding four “sides”).  Conversion and noise reduction proceeded nominally until I got to the fourth “side”:  It was backwards.  Reversing the audio was trivial, but figuring out why it had occurred on a cassette had me stumped for a while.  I eventually came up with a reasonable hypothesis.  But first, some background.

 

Both mono cassette and reel-to-reel recorders capture sound in the top half of the tape (Side A; the bottom half of the tape, where Side B is captured, becomes the top half when the tape is flipped over).  Stereo cassette recorders capture the left and right channels of Side A in the two top quarters of the tape; the third and fourths quarters are for the left and right channels of Side B (which become the top two quarters when the cassette is flipped over).  A major advantage of this approach is that stereo cassettes can be played successfully on mono players, and vice versa.

 

Stereo reel-to-reel recorders work differently.  Instead of using the top two quarters of the tape for the left and right channels, the left channel is recorded in the top quarter while the right channel is recorded in the third quarter.  (The second and fourth quarters become the first and third when the tape is turned over to record Side B.)  Because of this design, playing a mono reel-to-reel tape on a stereo player (or a stereo tape on a mono tape player) will play sound from one side as well as backwards sound from the other side (see Figure 1).

Mono tape vs. stereo R2R deck.jpg

Figure 1.  Mono open-reel tape as played on a stereo tape deck.

 

I expect backward content when I digitize mono reel-to-reel tapes using my stereo deck.  I turn the playback volume of the right channel all the way down so that I capture the desired content in the left channel and just a bit of noise in the right.  I then mute the right channel (which zeroes out the noise) and mix the left channel into the right.  With the same content in both channels, I end up with a mono audio file, as intended.

 

What I believe happened with the cassette in question is that the original recording was made on a mono reel-to-reel recorder (further evidenced by some speed variability).  At some later point, that material was recorded onto a cassette while being played on a stereo reel-to-reel deck.  This would result in normal audio in the left channel of both sides of the cassette but reversed audio in the right channels.  Still later, new material was recorded in both the left and right channels of the cassette’s Side A and just the left channel of Side B.  Side B’s right channel, the fourth I processed, remained backwards.

 

Content

Beyond technical aspects, personal recordings entail challenges arising from tapes’ content.  These are explored in the article Content Challenges Posed by Personal Recordings.

 

 

Back to Paul’s Blog & Contents

bottom of page