#76 Fractal Audio Axe-FX II teardown and repair

A couple weeks back, an acquaintance contacted me about his Axe-FX II suddenly not powering up anymore. I saw this as an opportunity to at the very least take a peek under the hood of one of these things (still selling for four-figure amounts), so why not?

I had neglected to snap photos of the outside of the thing (because other priorities), so you'll have to settle for some "stock" shots. Not that there's anything overly interesting to see anyway.

On to the more relevant(?) details, though. 

Nothing really shocking in there. Some analog stuff for the ins & outs, and a whole bunch of digital doohickeys, which is the whole point.

Moving swiftly on to the analog power section, pretty straightforward things - an LM317 and LM337 (both FairchildSemi, now OnSemi) regulating the positive & negative power rails for the analog stages. Nice touch with the 1N5352B 5-watt 15V zeners on the outputs of the regulators, for a bit of extra overvoltage-protection. Even though the mains power supply is rated to provide "only" +/-12V on the respective rails. But more on that later...

Next, an OnSemi MC78M05B is in charge of a 5V rail - most likely powering the converters. The mains power supply also provides a high-ish current 5V rail, but the 78xx series need the input to be at least 2V higher than their output, so this is likely powered from the +12V rail.

The power going to the digital side is a wee bit more... complex, let's call it. Powering what is most likely the bulk of things is a ready-made voltage regulator module, namely a Texas Instruments PTH05060W step-down regulator. According to the datasheet, and going by the two "1742"-marked resistors (17.4k) near it, this is set to put out 1.2V. Quite the price on this little bugger too, about $18 - wow... In the vicinity, a National Semiconductor / TI LM3965 powers the RAM (going by the "VDD_DRAM" label of a nearby test-point, at least).

Two National Semiconductor / TI LM1086 linear regulators provide two other supply rails for the digital side - my best guess would be 3.3V and perhaps 2.5V, since those would be less likely to require particularly high currents, given the context.

Moving on to the audio circuitry, most of the opamps are Analog Devices OP275's. Just for reference, they seem to retail for $1.8-$2, in bulk quantities (100pcs and up). That's worth keeping in mind, especially since there's more than a handful of them, strewn across the board. There's also an Analog Devices ADG212 quad switch in the input section, as well as an Analog Devices AD5290 10kohm digital potentiometer. Thank you, Google image search and LCSC for figuring out what the "4U" or "D4U" marking on an MSOP-10 package with the AD logo means...

The rest of the analog side is more of the same, OP275's and AD5290's. Fun fact - the latter seem to be well north of $3 a piece, even in bulk quantities. "By the way", the ADG212's are $2.5 and up, in bulk. Nice touch with the film caps, but still that's a loooooot of pricey silicon...

The conversion is handled by the nearly-ubiquitous Cirrus Logic CS4272. In the same area we can find a Cirrus Logic CS8427 digital audio transceiver. 

Now we're getting more into the "big guns" area, the guts of the digital side. Supervising the whole show, and likely handling the the front panel (dot-matrix display, buttons & lights) is a Microchip PIC18F83J11. Assisting it is a Xilinx Spartan-3 XC3S200 FPGA, while most of the firmware is stored in the neighbouring flash memories, an SST39VF1681, an SST39VF3201, and a Cypress Semiconductor CY62167DV30.

The USB interface is mediated by a Cypress Semiconductor CY7C68013A.

And at long last, the money shot (no pun intended..?) - the main muscle of the whole unit is a pair of Analog Devices ADSP-TS201S processors.

Assisting the DSP's are a pair of Micron MT48LC4M32B2 128Mbit / 16Mbyte RAM chips. Near the pedal inputs are an Analog Devices AD8541 single opamp, as well as a National Semi / Texas Instruments DS8921A differential line driver/receiver.

And now, to the "elephant in the room", as it were. The issue with the unit was that it no longer powered up. Allegedly even replacing the fuse hadn't helped. Upon plugging the thing into the mains, i tried to listen for any ticking sounds, which might've indicated some short-circuit on the output, and the power supply in the "hiccup mode", trying and failing to start. 

Not a few seconds later, i was surprised by a "pop!" and a really brief flash from under the mains heatsink. Great...

The basic design is nothing to write home about - your garden-variety flyback converter. The main controller was a UC3842, and a great example is found on page 22 of the datasheet. In this case, the list of parts that had in fact released the magic smoke was quite extensive. Namely, the MOSFET, the 1.3ohm current-sense resistor from the MOSFET's source pin, the resistor between that and the UC3842, the resistor in series with the gate of the MOSFET, the rectifier diode powering the UC3842, as well as the rectifier diode on the -12V output. As well as, "of course", the UC3842 itself, because why wouldn't it be toasted too?

A few afternoons of probing and checking later, i ended up concluding that the auxiliary transformer winding, the one powering the UC3842, must've shorted together. That was after replacing all the visibly fried parts, as well as the TL431 in the feedback network on the secondary side, the smaller electrolytics on the primary side, and the first one on the +5V rail, and measuring all the other diodes and resistors to make sure they were fine. 

The conclusion came from the fact that the power supply kept trying to start, with the voltage powering the PWM controller cyclically rising slowly, and quickly dropping as soon as it reached the starting-threshold (about 16V, in this case). The slow rise was explained from the charging of the filter capacitor through the 100k start-up resistor, fed from the rectified mains voltage. The shorted winding not being able to provide the voltage it was supposed to, left the UC3842 quickly draining its filter capacitor until its own undervoltage-threshold was reached, upon which it stopped; rinse and repeat.

 Once that was clear, i was utterly thrilled to see that the format (and even output connector pinout) of this power supply turned out to be somewhat of a standard. I fortunately quite quickly managed to find a Mean Well unit, a PT-45B, at one of my usual component retailers, which seemed to fit the bill well enough, AND was not a bottom-of-the-barrel unit. As opposed to the unknown manufacturer of this stock piece of junk, Mean Well are quite a reputable brand in the field, and they use Japanese electrolytic capacitors (for the most part, anyway), which gives them a leg up on longevity and reliability. And it didn't even cost all that much, to boot - about 23eu, including UPS shipping.

And now for the rant. Bear with me here - on a quick back-of-the-envelope calculation, there's about $70 worth of opamps and digital potentiometers on the analog side, and likely $1-200 worth of parts on the digital side. I understand, R&D has its costs, but these things retail(ed) for $1500-2000. And yet, they could't be arsed to shell out more than $5 for the cursed power supply? Seriously? But $18 for the little buck regulator to power the DSP's was just dandy. <sigh...> 

Some companies really have no shame... Neverminding the inevitable obsolescence of such devices, in the first place.