#34 MOTU Traveler mk3 teardown and repair

Hunting for faulty stuff on eBay has been fruitful yet again: this time, a MOTU Traveler mk3 Firewire audio interface. The reported issue was by now, the "classic", dead Firewire.

A pretty comprehensive little package, with loads of features shoehorned into quite a compact package - 1RU thickness / height, but only about 80% of the width.

Although truth be told, they DID cheat a tiny little bit - a few connections had to be moved to one of the sides.

But photos of the outsides are all over the Internet, as well as the manual. The arguably more "juicy" stuff is on the inside, right?

Lots of features in a small package, but they also had plenty of room to spare. A bit under half the depth of the enclosure is empty!

The brains of the operation is an Atmel (now Microchip) AT91SAM7S256AU, pretty beefy little thing. The USB connector is most likely for programming / debugging, and perhaps even controlling the ARM chip itself. I wasn't curious enough to plug in a cable - not sure i would've had the right drivers anyway. At least part of the firmware resides in the neighbouring AT45DB161D 16Mbit Flash chip.

You'll have to excuse the messy-looking TSB41AB2 - the one i had at hand might've come from some other unit, and i couldn't be bothered to clean off the flux. Works just fine, though. Considerably better than the stock one anyway (which had four of its eight data lines shorted to ground, and the other four showing a measly 0.11v reverse-drop from ground).

The "heavy lifting" (ADC data, DAC data, Firewire data and mostly everything in between) is done by a Xilinx Spartan-3E XC3S500E FPGA. Looks like it's still available, for $36 or so - not exactly "peanuts", i'm sure you'll agree.

A TI TMS320C6722 handles all the audio number-crunching - the mixing plus all the DSP equalization, dynamic compression, as well as reverb. The latter (for the most part) is the reason for the associated Micron 48LC4M16A2 64Mbit / 8Mbyte PC100/133 SDRAM chip.

The four microphone inputs are handled by a quartet of TI PGA2500 digitally-controlled preamplifiers. Not 100% sure what exactly the relays (Omron G6J-Y DPDT latching) are handling, but several possibilities come to mind. Either the pads (signal attenuation), or the phantom power, or the switching between microphone (XLR) and line-level (1/4" TRS) inputs - some combination of two of these functions. Tucked under the DB103G rectifier bridges (employed as signal clamps) are some opamps, likely the buffers for the line-level side of the combo inputs.

The outputs of the preamplifiers go into a bank of ST TS922I rail-to-rail opamps, and then on to a pair of Cirrus Logic CS5368 8-channel ADCs. Going by the number of opamps there, and the differential inputs of the ADCs, my guess is that the polarity-flipping (normal or reversed) of the eight analog inputs takes place here, and not in the digital domain. That would explain the placement of the two NXP 74HCT14D hex inverting Schmitt-triggers and the 74HC595D shift-register.

In the vicinity we also find an AKM AK4382 stereo DAC, dedicated to the headphone output, with a lone NJM4580 handing the low-pass filtering duties.

A single AKM AK4358 8-channel DAC handles the analog outputs; the low-pass filters / output buffers are hidden under the daughterboard carrying the top row of 1/4" jack sockets.

Speaking of which, said daughterboard contains line inputs 5-6, as well as line outputs 1-4. Not a lot going on here.

In the middle of the board, on the inside edge, there's the bulk of the power supply circuitry. From left to right, we have the boost-converter for the phantom-power (given away by the factory-bodge 63v capacitor), complete with a single-stage Cockroft-Walton voltage doubler; the +/-5.7v supply, for the opamps; and the 3.3v supply (silkscreen designation is behind the big fat electrolytic cap).

Speaking of which - a complement of Meritek and Lelon electrolytics. I respectfully decline comment (but feel free to read between the lines)...

The 3.3v supply is handled by a TI TPS5430 step-down converter, same as the analog supply. Unfortunately i didn't write down the markings from the boost-converter, and i can't quite make them out from the photos.

The last(?) thing of note is an almost-hidden TI TLC2933A PLL - likely for handling the clock synchronization for the digital I/O.

Now, last year i had gotten my hands on a similar unit, but that also had a pretty well-"baked" power supply. In order to avoid issues down the line, i figured i'd crack this one open (somewhat literally, as well), and do a preemptive re-cap.

Rated at 15v @ 1A, it doesn't look too shabby. It actually has some input filtering (X2 cap plus common-mode choke), and some proper silicone caulk to mitigate vibration-induced damage. At least it's not the infamous and well-deservedly dreaded "brown glue" (which, in addition to being corrosive, it turns black, brittle and conductive with age and heat).

Since there's no controller chip to be seen, only a TO-220 packaged high-voltage MOSFET (couldn't make out the markings), and two other TO-92 devices on the primary, means it's a self-oscillating flyback circuit. On the bright side, there's optocoupler-feedback from the secondary, which is always welcome (as opposed to just inferring the load from an auxiliary primary winding).

One of the TO-92 devices is a 2SA1015 PNP transistor, but no idea about the one behind it (and i wasn't curious enough to de-caulk it to find out).

The rectified-mains bulk filter cap is a Teapo - it gets a passing grade in this low-stress role. At least it's not some bottom-of-the-barrel no-name brand.

Secondary side though, "SWC" - whole lotta nope! One of those had failed so bad in the previously-mentioned case, that it had blown out the rubber bung on the bottom. This and its twin got replaced with a pair of Rubycon ZL 330u/25v caps, plus a Panasonic NHG 1u/50v across the TL431 in the feedback circuitry, and the 4.7u/50v handling the "aux primary supply" filtering duties made room for a Panasonic FR 18u/63v.