Introduction
End of 2017 I purchased a good-condition Nakamichi Cassette Deck 1 (aka
CD1), for no other reason than that I lusted after such machines back
when I could not remotely afford them. I also always wanted to learn the
real performance of the cassettes I so lovingly used in the 80s and 90s
(first with a TEAC V-455X, then with a Nakamichi CR-2E), until
CD-R and self-recorded DVD-A erased my interest in analogue recording. So in 2018 I
embarked on a project of characterising all tapes I could lay my hands
on. I did so first using that CD1, but after a while I changed over to a
Nakamichi BX-300E, and I improved my measurement methods.
The photographs are of the actual tapes used for the test, all sourced
in Europe with a few Japanese exceptions. Many of these cassettes are very old now. Initially most were mine,
and had been well taken care of. Later I started purchasing used cassettes in Belgium, Netherlands, Germany, and Slovenia, to fill the gaps in the collection. These used tapes may have some wear. Finally, with these pages gaining some popularity, people started donating me tapes. U No Who U R. Many thanks!
Machinery
I have one deck dedicated to measurements only: a refurbished and slightly modified Nakamichi BX-300
with no detectable head wear.
This BX-300 can record quite high levels
onto type I and type II tapes without much distortion, making the deck
suitable for exploring a tape's maximum output level (MOL). The playback
response of this deck, checked with TEAC MTT-356 and ABEX TCC-260 IEC
1981 tapes, is flat within 1.5dB between 200Hz and 10kHz, slightly
rising thereafter, as is normal for a narrow-gap head. Moreover, the particular frequencies used in my tape
evaluations, 400Hz and 10kHz, replay within 0.5dB of each other. With
its bias aligned to a IEC-compliant tape, such as some iterations of TDK
SA, a very flat and extended frequency response ensues. This shows
that the BX-300 is a valid platform for tape evaluation, devoid of
idiosyncratic record or playback equalisation. I have a full set of
alignment and calibration tapes (TEAC, ABEX, BASF, Roth, and A.N.T.), and
regularly check this deck for accuracy.
Signal generation and
capture is done with an Alva Nanoface USB sound card running at
96kHz, using the excellent audioTester
software. The frequency sweeps are made with 190 milliseconds
compensation for the latency between record head and play head. (Since Summer 2022 I am using the MOTU M2. It is easier to configure, much quieter, and has a higher input impedance. Its only drawback is a slight premature bass rolloff on the line inputs. This is compensated for in audioTester. Most cassette tests were done before 2022, though, thus still with the Alva.)
Initial setup
For each tape under test first the bias level is determined by making a frequency sweep at -20dB (relative to 200nWb/m ANSI), tuning the deck's front-mounted bias pot to obtain the flattest response below 10 kHz, while still reaching beyond 20kHz, and with up to 3dB peaking accepted above 10kHz.This is obviously a somewhat subjective process, and not entirely repeatable: when is flat enough flat enough? When is a peak acceptable? Further, a tape's bias requirement may vary somewhat along the full length, especially when the tape has accrued significant wear, as was evident in a number of cases. The bias reported in the test results is as read off the BX-300's adjust knob scale, which runs from -5 to +5: '-' or counterclockwise corresponds to reduced bias current and thus increased treble levels, '+' or clockwise to increased bias current and reduced treble levels.
Measurements
After setting bias the level-related tests are done (sensitivity, MOL, SOL, noise, see below). The frequency sweeps are then repeated at -20dB, -10dB, 0dB, and +6dB. Due to treble saturation the bandwidth shrinks with each amplitude step. This is normally the worst with ferro-cobalt and chrome type II tapes (due to the higher treble boost of the 70us recording equalisation), and the best with type IV metals. Even so this treble loss is relatively harmless, since the natural spectrum of most music also rolls off with frequency.
The +6dB line often lies less than exactly 6dB above the 0dB line. This is due to compression: the tape progressively returns less signal with increased input. Note that with tapes having a MOL much less than 5dB, the 6dB line is recorded with gross distortion, and thus is not representative at all of real use.
Sensitivity is measured at 400Hz, at -3dB relative to Dolby level for type I and II, and at 0dB for metal. Sensitivity is not a quality parameter, but an indicator of compatibility. When a tape's sensitivity differs from the exact tape the deck was aligned to Dolby noise reduction will mistrack, resulting in reduced or excessive treble and/or audible pumping and breathing.
Maximum Output Level is measured off-tape, relative to Dolby level, at 400Hz. MOL is determined for 3% Third Harmonic Distortion , as per the industry standard, and also for 1% THD. The difference between the two levels gives an idea of how fast distortion rises with increased input signal. A.N.T. Audio has an interesting collection of distortion-versus-level sweeps for tens of tapes, clearly showing their differences in distortion signature: HD3 v level at 315 Hz for 104 different cassette tapes.
3% THD MOL is also measured at 1kHz. MOL (and SOL for high frequencies) gives an idea of how loud the tape can be recorded safely. This is necessary to exploit its available dynamic range, but it is important to know that many decks are not capable of driving such high levels onto tape.This is one aspect in which the Nakamichi discrete heads excel.
On a good deck dynamic music can be recorded with the peaks at the MOL400(3%) level. Less dynamic, compressed music, or music with a lot of treble or a lot of loud deep bass, can better be recorded to the MOL400(1%) level. When noise reduction such as Dolby B or dbx is used, it is better not to record beyond 0dB at all: the resultant loss in treble extension due to saturation will be multiplied by the noise reduction scheme, while the gain in dynamic range thanks to the high recording level will be insignificant compared to what the active noise reduction already is doing. (Dolby C has an anti-saturation network, avoiding this pitfall. But again, why would you hammer a tape for a relatively small gain in dynamic range?)
Saturated Output Level is estimated by increasing a 10kHz input signal until no further rise in output level can be seen. This is a very coarse method, since the output-versus-input curve for most tapes is very shallow at this frequency. It is also a bad metric, as at saturation the signal already is very heavily compressed! A better method would be twin-tone intermodulation distortion, but so far I have not succeeded in getting plausible results with my setup.
Finally bias noise is measured, by recording silence and reading the resulting averaged RMS signal level integrated over a 20Hz-20kHz bandwidth, then subtracting this from Dolby level. This is done unweighted and also A-weighted. One single stereo track (0.6mm) is used. (Some tape datasheets in the past quantified noise for a 1.5mm mono track, resulting in a 4dB better figure.)
As per industry practice the tape's dynamic range in dB is then found by subtracting the A-weighted bias noise from the MOL. The innate noise of the BX-300, i.e. playing without tape, is -57.3dB or -64.7dB(A) with 70us equalisation, still a few dB quieter than even the best tape.
All levels are relative to Dolby level, which is defined as 200nWb/m ANSI, equivalent to 218nWb/m DIN. If you want to translate to IEC or DIN zero level (250nWb/m), a reference level used on some more recent decks, then simply subtract 1.2dB from MOL, SOL, and noise levels. For example MOL400 = +5dB becomes +3.8dB, and noise = -46dB becomes -47.2dB. Dynamic range, obviously, remains the same.
Note that older decks had their zero level at 160nWb/m (DIN), requiring the addition of 2.7dB to the MOL, SOL, and noise levels of my reports for proper translation.
Validity
The set of parameters shown here is of course not complete. There is much more to tape quality than bandwidth and dynamic range. One must also consider print-through, uniformity, drop-outs, azimuth accuracy, modulation noise, wow&flutter, consistency, heat resistance, and mechanical stability and reliability. From past reviews we know that a cassette excelling in one corner (e.g. exceptional dynamic range) may well drop the ball in another corner (e.g. print-through).And even all the standard measurements in the world would not describe how a cassette would sound. A tape acts as a level-dependent and frequency-dependent compressor. If we were to make frequency plots at all levels, from -40dB to +10dB, and if we were to measure harmonic and intermodulation distortion at all levels and at all frequencies, then no two tapes would show the same results. They would all have their own distinctive fingerprint-like patterns.
Also, the results here are no absolutes: a first reason is that a tape cannot be analysed outside of a deck. The magneto-acoustical parameters (sensitivity, MOL, SOL) are dependent on the amount of applied bias. Yet, a deck is designed so that the combination of its median (a.k.a. standing) bias and its particular recording equalisation yields a flat frequency response. The designers can select the standing bias according to their own criteria, and then build the entire record path around this. Most decks will attempt to strike a balance between MOL, SOL and bandwidth, but there are also decks that prioritise MOL, at the cost of bandwidth, or vice versa. Such decks likely will obtain quite different results from a given tape. Still, my results should be broadly comparable to those obtained by audio magazines, especially the ones that used standard-issue Nakamichi decks aligned to the IEC references.
A second reason is that the typical TDK SA or the typical Maxell XLII does not exist. Within manufacturing batches there is sample to sample variation. There is variation between batches. And finally there is evolution: ostensibly identical instances of a particular tape may well be different due to improvements - or cost savings - implemented during its commercial life cycle, without informing the public. Variations of +/-0.5dB on key parameters, and even more, are to be expected. Due to this my results may differ significantly from yours or from old magazine reviews.
Dating
The photographs are of the actual tapes. Dating is done based on http://vintagecassettes.com/, with a second opinion from http://www.45spaces.com/audio-compact-cassette-blank-tapes and https://www.ez647.sk/cc/audio.html#typII when needed. These sources for dates are not entirely without error, though.
About these pages
The layout of these pages may appear a bit inconsistent or even sloppy. That is mostly due to issues with Google's blog editor and the rendering in your browser: you never get quite the same result twice, WYSINWYG, and just saving and re-opening a page in the editor is sufficient to mess things up.
Some frequency curves show a quick drop off in the deep bass or the treble. These are artefacts from audioTester when at the start or end of a sweep, and have nothing to do with the true performance of the deck.
Some frequency curves start at 200Hz instead of 20Hz. These are older tests, shortened to save some time.
Finally these pages were never meant to be an actual blog: I just wanted a platform for quickly publishing these notes for myself, accessible from any device at any place. I decided to make this all public only afterwards.
Example: Maxell XLII (1994)
Relative bias: (reference)
Relative sensitivity: (reference)
THD @ Dolby level: 0.62%
MOL400(THD=1%) : +1.8dB
MOL400(THD=3%): +5.3dB
MOL1k(THD=3%): +5.5dB
SOL10k: -3.4dB
Bias noise: -54.8dB, -59.2dB(A)
Dynamic range: 64.5dB
INDEX OF ALL CASSETTES
