It is probably a good moment to announce this important information: after Model S USB I don’t plan to design D/A converters based on TDA1541A anymore.
And, even though I wasn’t really dogmatic or exclusive about D/A chips – in fact, I designed different converters for other manufacturers in the meantime, and each of them was using a different D/A chip – or dogmatic at the general level for that matter, I am aware of the number of people who associate or even identify my work exactly with TDA1541A.
Hence I also find it important to stress that the reason behind this decision was not any real shortcoming of the chip itself, but its true unavailability in the longer term. In fact, the same reason already forced me to abandon TDA1541A in the lower priced DACs. (For more details on TDA1541A availability, please see part 4 – TDA1541A grades and series.)
Obviously, my decision to use TDA1541A as a reference all this time, including the recently released reference USB converter Model S USB, is not in agreement with criticism regarding the technical obsolescence of this (or any older) chip. In addition, such a presumption of obsolescence is often associated with the spread of prejudice about non-oversampling DACs which “don’t measure well, even if they sound fine”.
So, what’s the verdict on TDA1541A? Why and how is it still possible to make a reference D/A converter based on the chip that was first introduced more than 25 years ago?
First of all, it is important to understand the reasons behind the early 90s’ semiconductors industry move from classic multibit (often referred to as “R2R”, even though the real R2R architecture was abandoned quite early) to delta/sigma converters, and to realize that these reasons were practical in nature, and cost related. Traditional requirement for highly precise and thermally stable string of current sources, with delta/sigma moves to the domain of digital processing, which significantly decreases the costs. This process shifts a multibit PCM stream to one bit stream at a higher sampling frequency, usually known as PWM or PDM (there is a slight difference between PWM and PDM but we can put it aside for now), and which is an equivalent of class D (“digital”) amplifiers’ output. Such converters, of course, employ the most simple and straightforward output, and it is usually already voltage output, so the one that doesn’t even require an external I/V stage. Still, modern delta/sigma DACs are not one bit but use several bits output, up to five or six. Consequently, their work is then to process 16 or 24 bit input into 5 or 6 bit output. Such a 5 or 6 bit signal is still relatively undemanding for conversion, yet it overcomes some inherent problems of real one bit process, such as modulation noise and jitter sensitivity.
For some of us however multibit chips remained a reference for their natural tone and timbre. There is no doubt that further delta/sigma developments are possible too – late research reveals some previously hidden problems associated with this process – but what matters here is this: there is a primary difference between two converter types, both of them having their own story.
And the classic multibit D/A chips story mostly stopped in the 90s. Philips in fact stopped in the 80s with TDA1541A, Analog Devices stopped with AD1862, and the only remaining multibit representatives are Texas Instruments (Burr-Brown) PCM1702/1704, which were also designed back in the 90s. Each of these chips should be considered not one simple point in DAC history, but the end and crown of its own era.
My bet was TDA1541A, for its romantic tone and unbeatable dynamics. Also, TDA1541A in my view never fit well tube output stages, and I believe that my output stages are way better match, especially for their sonic features associated with the absence of feedback at one, and exceptional linearity at the other side. And once I found myself in this, I also understood how working with the same part in the longer term is beneficial: you realize its properties, needs and requirements, and thus the best way to use it.