Eisen Audio

AWB95 Cassette Tone Colour PCB Kit (Discontinued)

Eisen Audio's AWB95 Colour lends the mid-heavy crunch of lo-fi, cassette electronics with the dark, wooly sound of worn-out tape. The AWB is sold as a partial kit, including the PCB, Colour connector, and nylon standoffs. The user must source the rest of the parts from the Bill of Materials.

Note: As will all our PCB projects, it's a good idea to check the availability of parts in the Bill of Materials before ordering.


  • NOS opamps
  • Partial kit comes with PCB, 8-pin Colour connector, and nylon standoffs
  • Transformers recreate magnetic hysteresis of tape
  • Compatible with the Colour format
  • Black PCB
  • Full, step-by-step assembly instructions

Notes on the Sound and Circuit

In designing the TM79 Colour I touched upon the elusive challenge of tape emulation circuitry, ultimately arriving at a refined sound that is not so obvious. For my second saturation Colour, AWB95, I focused on a more exaggerated lo-fi tape sound. Initially I had in mind my impression from the first time I put up a reel of GP9 on a tired Tascam 52 at 7.5IPS and transferred over a CD. The resulting sound quality was big and dark and slow: “low speed." Pondering the low speed theme (which is what an earlier version of this AWB95 Colour was tentatively titled) while continuing to tweak my circuit, I was met half-way by a phenomenon which has become my underlying philosophy of creative design: you start out thinking you know what you want the thing to be, but in the process of manifesting your vision the thing shows you what it wants to be. Indeed, after a long night of auditioning my low-speed circuit it was sounding less like 1/4“ 7.5IPS and more like an even slower speed consumer cassette tape. Feeling inspired, I started chasing the sounds I remembered from even earlier audio experiences of my adolescent years: all that time spent listening to an AIWA walkman, a AIWA boombox, and a Ross cassette 4-track. Revision C of these experiments is the AWB95 Colour which brings a smile to my face every time it’s engaged, and I hope you will have the same enjoyment. The circuit is described as follows.

The input coupling cap, C5, is a Nichicon FG series which I’ve used before for its unique tonal signature that I’d describe as “heavily damped, wooden, unfinished” which I suspect comes from its dielectric absorption and associated parasitics. At 3.3uF it is smaller than I’d normally choose (less energy storage, less bass) and I’ve provided no direct drain path so that it’ll “stay full”. From here we have two parallel paths for the signal. 

First is the active path where a 3M3 series resistor and the aforementioned lack of immediate DC drain path aim to underbias the IC op amp, U1, thereby increasing THD+N, while forming an RC lowpass filter with the op amp's input capacitance which makes transients and very high freqs appear "softer." The op amp’s DC-coupled load is an undersized 600Ω repeater coil, T1, which doesn’t really have enough self inductance to pass real bass at the intended operating impedance. Indeed, this is a telecom transformer only rated down to 300Hz. I am using both sections of U1, an RC4558 dual op amp, in parallel via the 10Ω combining resistors in order to drive T1 as efficiently as practical without resorting to a higher powered amplifier stage (i.e. discrete transistor follower). Furthermore, it is not loaded with 600Ω at low frequencies, and as a net result we maintain some musical low end through T1 while still getting the premature distortion and saturation intended by choosing an undersized transformer. Some of this perceived compression is guaranteed by the secondary termination resistor, R6. Because even tiny transformers affect high frequencies less than lows, and because we want obvious coloration throughout the audio spectrum, I’ve included a symmetrical diode clipping network, D1 and D2, which take affect only after the network of R7 and C7. This means frequencies 1.3kHz and above - the second half of midrange and HF - encounter the  double loading of R7 followed by the passive limiting of D1/D2. C3 is a WIMA film capacitor because of its perceived damping and HF texture. D1/D2 are 1N34A germanium because I have observed that, in addition to the lower ON voltage of Ge, the clipping behavior of germanium has a very soft knee, and particular junction capacitance, and a duty cycle modulating behavior which combine for a clipping behavior that seems “darker” and more “tape like” to me than what silicon diodes can offer. Next we introduce a similar but larger transformer, T2, which deals with an interesting set of contradictions. Because our signal path thus far has been deliberately lossy we would like some makeup gain, and T2’s slight set up ratio should provide approx. 4dB of it. However, because T2 is not driven directly from an efficient low-Z source but rather via another transformer, because its own coupling is not great, and because of its turns ratio, T2 presents something of a source impedance to the outside world, which is further built out by the series L1 and R5, forming a voltage divider with the 10kΩ load of the following circuit stage which ultimately results in a bit more loss. Although T2 has a larger core with greater self-inductance and power handling with which to pass a fuller audio bandwidth, it is still a cheap telecom transformer less efficient than what we’re used to in pro audio, and so it too has a harmonic contribution. The aforementioned inductor, L1, is an RF coil meant to impede only transient response and high frequencies, once again attempting to compensate for the transformer inherently affecting low frequencies more heavily.

A version of the aforementioned signal path was initially my complete AWB95 circuit, and I was quite satisfied with the lo-fi charm of its midrange saturation and high end rolloff. However, upon evaluating an early prototype Peterson complained of the bottom end rolloff and I was forced to admit that, even with “cassette sound” as the end goal, this was unacceptable. Correcting for the LF response lead to the parallel passive signal path of R8, L2, and C4. L2 is a 220uH power inductor in series, meant to rolloff high frequencies while passing only sub bass with some added resonance and saturation. C4 makes sure very high frequencies are rolled off so as not to trump the subdued transient response of the primary signal path. R8 serves two purposes: first to damp L2 and limit its Q (limit the resonant bump) by raising source impedance, and second to provide loss as this parallel path passive combines with the output of the primary active path, thereby helping to set bass amount and overall gain.

As touched upon, even though both signal paths are lossy for effect, the AWB95 Colour manages to maintain a sense of unity gain via the step up of T2, the resonance of L2, and the overall THD+N.

As with each individual component of the AWB95, the NJM4558D dual IC op amp was specifically chosen by ear. It’s particular brand/model should not be taken for granted and not substituted, because not all 4558s are created equal. For example, there is a NOS military CDIP version that I use in the NonLinearAudio CurvOmatic filter and tone control product because out of 60 ICs auditioned, it demonstrated the most extraordinary bass resonance. On the contrary, this Japan Radio 4558 sounds pretty “shitty,” imparting a certain midrange smear that I find nostalgic. Having seen this cheap chip in so much lower priced gear from the 70s and 80s, and recognizing its particular tone, I suspect something similar (such as the JRC4565D in DIYRE's Distortastudio Colour) was used in the consumer devices which inspired this AWB95 Colour.

-Jens Jungkurth, Principal, Eisen Audio

Related Products