Assemblage Discrete I/V Mod

Friends, the year is 2008, and discrete I/V mods are in. I had some spare I/V boards from a DIY PCM1798 project and decided to throw one into an Assemblage DAC-2 that would occasionally cut out.

The Assemblage, for those of you who do not follow vintage DACs, was a pretty interesting value at a time when people were turning to PCM1794s and Wolfson for endgame sound. Strange times. Mr. Sneis has a good story about these. This one uses the well regarded PMD100 HDCD filter and a pair of PCM1702s.

And that's about all it's got going for it. The whole process unearthed so many strange and questionable design choices. Without wanting to go full KG, here's an illustrated guide (photo from an old FS).

• The transformer generates +/-25V supplies, that are fed to a bunch of regulators near the center of the board.
• The analog section is powered from some very fancy LT regulators.
• This is then regulated down to 5V with some 7805s for the HDCD and CS chips. I would not power tube heaters with these, but surely it's fine for digital?
• The DAC chips are fed from the 7805 and 7905 on the bottom left. The 7905 is on a heatsink, because the PCM1702 draws much more current from the negative rail than from the positive rail.
• There is a big cement resistor below all the regulators that gets very hot. Its true purpose is a mystery, but the leading hypothesis is that this is a dropping resistor for the input selector relays.
• The output relays are controlled by a darlington array chip up next to the HDCD chip, using the transistors as switches. The input selectors are controlled by using the switches on the front panel as switches that control the relays, which then do the actual signal switching.
• The I/V and LPF section are reasonably conventional. Of note are the many expensive Multicaps and Caddock resistors. Some of the resistors were not actually soldered in, and this is probably why the signal would cut out. A few of them are used in the relay circuit.

So, the power supplies will definitely need to go, but that is a project for another weekend. I pulled everything between the DAC and ouput jacks. Here's what's going in its place.



It's a very simple but good sounding circuit, similar to Jocko's I/V. There are plenty of articles about the virtues of discrete I/V stages, but one of the key benefits is that, unlike op amp I/Vs, it does not rely on feedback to lower the input impedance. Op amps have tons of gain at DC, but it tends to fall off pretty early (as low as 100Hz, for something like an OPA637), and so an op amp I/V's input impedance increases with frequency.

The input impedance of a common base BJT depends chiefly on the current running through it, and on the impedance at its base. This holds true until a few MHz. So, the secret sauce is to run it hot. At 10mA, it has an input impedance of about 2.6 ohms. (Folding the cascode, as this cetoole fellow once did, would allow higher currents and DC coupling with the same PS voltage.)

The PSRR of this circuit is basically zero. While it might be tempting to put a tube rectifier in the power supply, I opted for simple shunt regs to keep any big caps out of the output current loop. It's quiet.

Here's the DAC with the I/V board installed, and some caps replaced.