The reel project aims at improving analog tape simulators in software, and tries to discover whether it is feasible to develop a simulator that really closely mimics the ‘reel thing’.
Analog tape frequency response
As discussed in the previous post on magnetic tape recording, we’ll have a closer look at the frequency response of tape recordings. For now we only consider the linear part of the tape recording transfer function, which can be measured with maximum-length sequences (MLS) or chirped tones. All measurements discussed here were done at a signal level of -5 dB on the device’s VU meters at 1 kHz to ensure that we have a proper signal-to-tape noise ratio, and a fairly linear behavior of the tape and the recorder as well. The recordings were made with a brand-new, empty RMG 911 Studio Master tape.
Let’s start with the famous Studer Revox A77 mk4. This tape recorder was produced from 1974 until 1977, and we are using the High Speed (HS) version that supports tape speeds up to 15 IPS. The frequency response for 7.5 (in blue) and 15 IPS (red line) is shown in the figure below.
Here we see some interesting effects of tape speed. Firstly, the frequency response is markedly better above 10kHz for a tape speed of 15 IPS. Furthermore, for 15 IPS, we see a bump at around 50 Hz, which is most likely the so-called ‘head bump’ that results from an interaction of the signal frequency and the construction of the read head. The head bump does not appear at 7.5 IPS.
For both tape speeds, the transfer function has a slightly negative slope, potentially contributing to the Revox’s ‘warm’ sound. Interestingly, when we compare these results with the frequency response of the Teac A4300SX, we observe a completely different frequency response, even though the same tape and tape speed are used.
This figure demonstrates the frequency transfer function for the RMG 911 Studio Master tape running at 7.5 IPS on the Teac A4300SX (red curve) and the Studer Revox A77 mk4 (blue curve). The former has a substantial gain in the high frequency area, as well as low-frequency head-bump effects. Both of these properties are absent in the Revox transfer function. Measurements of the Teac A4300SX at 3.75 IPS (not shown here) indicate a much flatter frequency response, suggesting that this unit is tuned to 3.75 IPS rather than 7.5 IPS.
From these simple examples it is obvious that analog tape recording can cause substantial differences in the audio signal spectrum, even at moderate signal levels that cause little saturation. Moreover, the employed recording unit seems to have a significant effect on the signal spectrum besides the tape speed, and hence a proper tape simulation algorithm should take these effects into account.
The two audio demos below clearly indicate the different character of the two recorders. A drumloop was recorded at the exact same (moderately high) level on both recorders at a tape speed of 7.5 IPS. The Teac A4300SX sounds quite bright, and handles the high signal level with relatively little distortion of the kick drum, while the opposite is true for the Revox A77.
In the next post, we’ll dive deeper into the subject of tape non-linearities and provide some early simulation results.