In the mid 1990s, MacroSystem Germany released the Maestro Professional sound card for the Amiga. It was a special sound card because it was fully digital, having only optical and coaxial digital connectors. It was suited for lossless recording from CD and DAT, as well as generating lossless audio output for DAT recordings. With tools like Samplitude, the Amiga became a studio quality digital audio workstation. There was also a tool for doing backups on DAT. At that time, these tapes were the cheapest way to backup entire harddisks (a 90 minutes DAT tape could backup almost 1 GB of data, which was a lot in the 1990s).

Unfortunately MacroSystem had never released a driver for the sound card, so it could only be used by a few (and mostly commerical) tools. I pestered their developers at every Amiga fair I could attend, but to no avail. Then, at the end of 1994, I decided to find the datasheets of the Yamaha chips, reverse engineer the board design, and write a driver myself. It took some time of trial and error, but eventually I was successful. In the coming years, my driver, the maestix.library (source code), became the inofficial standard driver. OctaMed Professional is maybe the most prominent software using it. Some professional music artists used Amiga and OctaMED for their production, so maybe my driver was even used for recording the masters of some famous CDs? 😁

Digital Audio in a Nutshell

The MaestroPro is able to receive and transmit digital audio data, either in the S/P-DIF or AES-EBU standard. The former one is still widely used in home equipment today, while the latter one was rather common in studio equipment. Today's standards permit different encodings and high sampling rates, but the MaestroPro could only read 2-channel 16-bit raw audio with sampling rates of either 48kHz (DAT), 44.1kHz (CD), or 32kHz (DAB).

Besides the raw audio data, the standard also transports Channel Status Bits (CSB) and User Data Bits (UDB). The CSB contain information like the used sampling rate and the copy prohibition state. The UDB are not standardized, and usually transport proprietary data between studio equipment.

Inside the Maestro

The board's design is straightforward. It mainly contains a transmitter, a receiver, and FIFO memory for transporting the samples between the board and AmigaOS.

The optical and coaxial inputs go to a Yamaha YM3623B Digital Audio Interface Receiver (DIR). This chip decodes the audio data stream, extracts the CSB and UDB, and generates a raw bit stream of the audio samples. Shift registers convert it to a 16 bit parallel stream, which is stored in a 1K x 16 bit receiver FIFO. As soon as the FIFO is half filled, an interrupt is raised, and the Amiga driver reads the received data from the FIFO. This happens up to 190 times per second.

The most important CSB are readable via a status register of the board controller. The UDB are copied to a separate 8 bit shift register, which could be polled by the driver. However, UDB are usually 32 bit wide, so reading them was never really used in practice (at least not to my knowledge). The Maestix driver only provided a very rudimentary API for the UDB.

On the transmitter side, the 16 bit samples are pushed to a transmitter FIFO, and then converted to a serial bit stream by shift registers. A Yamaha YM3437C Digital Audio Interface Transmitter (DIT2) converts it to a digital audio stream and sends it via an optical output. The Maestro Pro does not have a coaxial output, presumably because there was not enough space on the board for a fourth connector.

The DIT2 is unable to generate the sampling rate clock by itself. It needs an external clock source instead. On the Maestro Pro, this clock is generated by the DIR. It is either derived from the bit stream of the selected input, or generated by an internal fixed 48kHz clock source. For this reason, the Maestro Pro needs to rely on external signal sources for 32kHz and 44.1kHz output sampling rates.

The transmitter can choose from two data sources. One source is the transmitter FIFO. The other source is the bit stream from the DIR, bypassing the FIFOs. This enables the board to modify the UDB and CSB of the incoming signal directly, without involving the CPU. But since the transmitter and reciver paths are fully separate, the MaestroPro is even capable of providing full-duplex audio streaming. The maestix.library takes advantage of that with the "realtime FX" feature, where the signal is read from the receiver FIFO, modified by the CPU, and then immediately sent back to the transmitter FIFO.

The entire board is controlled by three GALs and a small handful of 74LS logic chips. They take care of the Zorro bus protocol, provide mode and state registers, and orchestrate the transmitter and receiver paths.

Broken MaestroPro

All of the components of a MaestroPro can still be found on the market, although both Yamaha chips are not produced any more and can only be found on some Chinese online markets as NOS parts. But basically, it is still possible to repair a broken MaestroPro.

The major weakness are the three custom programmed GALs. The GAL manufacturer states a memory retention time of about 20 years. It sounds like pretty much, but remember that these boards are almost 30 years old now. We already exceeded that life span by 50%!

When I reactivated my Amiga in 2021, my MaestroPro was working fine for a couple of minutes, but then it started to lose synchronization with the audio source. The only way to fix that problem was to turn off the Amiga and let it cool down for several minutes. A deeper diagnostics showed that the card seemed to detach itself from the Zorro bus. It seemed that one of the GAL chips had thermal problems, or was maybe starting to "forget" its programming. Fortunately I was able to recover the programming scheme. I replaced the original GALs with brand new Atmel ATF16V8C-7PU ones, and to my relief, my MaestroPro is now working stable again.

The fusemaps are copyrighted by MacroSystem, so I am not permitted to share them to the public. However, if you happen to have a broken Maestro Pro, please get in contact with me. Maybe I can help you to repair it.

The Maestro (without Pro)

There was a predecessor of this board. It was just called "Maestro", and had some major drawbacks. First of all, it had no transmitter and could only receive audio data. Secondly, it did not have a FIFO, so the sample words had to be read by the CPU as soon as they became available, which is up to 96,000 times per second. This was only possible by turning off multitasking and interrupts during recording, which also meant that recordings could not be written to harddisk, but had to be stored in RAM first.

Compared to its successor, the Maestro hasn't been a great success. I haven't seen one since the end of the 1990s, and I also don't know a single software that is actually using it. Due to the technical limitations, the Maestix driver won't support it.