#36 TC Electronic Digital Konnekt x32 teardown and repair

We're having a "productive" period, as it turns out. This time around a TC Electronic Digital Konnekt x32 audio interface / digital patchbay is on the chopping block. Or on the operating table, rather - i do indeed hope to repair, rather than dismember and mutilate this thing.

I managed to catch this, along with a twin of its, for under $200 / 200e delivered, as a "job lot", both sold as untested / faulty. The first one was shown powered up and possibly working fine (several LEDs on the front panel being lit), so potentially fine, pending tests. This second one though, not quite so much - only the blue LED halo surrounding the power button lit.


Now, these x32's are from the same family and generation as my "battle-scarred" Konnekt 48, and a photo of the internals that i stumbled upon online semi-confirmed a hunch - that they actually shared a power supply. Stock, my Konnekt 48 had Hermei electrolytics on the secondary side of said power supply which are... let's just say, quite far from the most trustworthy brand.

Also, in my K48, the power-button blue LED is driven from a microcontroller powered from a 5v rail, while the main MCU runs from the 3.3v rail - the former may still be marginally working, but no (stable) 3.3v, no booting or other (inter)action. That had been the case on a(nother) K48 i had bought as dead a year or two ago, and swiftly revived by replacing one single capacitor. Hence the hopeful confidence, in this case.

Quite the lineup of ins and outs - nearly all digital, in case the model name wasn't enough of a giveaway. The sole exceptions are a stereo pair of balanced line outs on the back (labeled "Monitor"), and a headphone output on the front - the level of both being controlled by the front-panel volume knob. Metal-shell Neutrik XLR's for the AES/EBU connections - you don't see that too often.

Slight side-note: the first unit did indeed turn out to be absolutely fine. The driver / control panel recognized it immediately upon connection to a computer (the Firewire connection could've been shot, but wasn't), clock setting and slaving worked fine, ADAT and S/PDIF comms with an old MOTU 828mk2 went off without a hitch. That on its own will pay for the whole package AND leave a bit of profit on top.

Right, let's rip straight into this thing... I won't lie, bit of a hassle, what with all the connectors being fastened to the rear panel (same piece of sheet metal as the top cover). Nothing a cordless drill and a PH1 bit can't handle, though.

Aaaaaand we're in. Not exactly fully-packed in there, but then again, you don't need a heck of a lot of circuitry to just marshall around streams of bits. Minimal analog stuff cuts down on the needed real-estate as well.

Right in the middle we have the already familiar Dice II Std running the whole show, along with the Samsung K4S641632K 64Mbit (8Mbyte) RAM and Spansion S29AL016D70 16Mbit (2Mbyte) parallel flash (where the firmware resides). Speaking of which, it kiiiiinda-sorta looks like the soldering on the flash chip's pins isn't factory, but maybe it's just my imagination.

Over on the left side it gets a wee bit more interesting - a pair of Texas Instruments / Burr-Brown SRC4184I four-channel sample rate converters. Right next to them, a Xilinx Spartan-3E XC3S100E FPGA. My guess would be, that's the one that does much of the so-called "glue logic". The Dice II is well capable of handing the format conversions on its own (a "measly" 16 channels), so all that's left is perhaps the actual routing between ins and outs, and controlling the SRC's.

On the right, we have most of the power supply sections, including the mains input. Ah, the mains power supply...

Sure enough, it's identical with the one in the Konnekt 48 - not exactly a major shocker. And yes, here's your cue to shudder... (Had to mess with the contrast & gamma, the markings are VERY faint)

There's the starting point for the repair, i'd reckon. Sure, step one is upholding the first commandment of troubleshooting - Thou Shalt Measure Voltages. All in due time though, first to complete the hardware "inventory".

Still quite baffling how the power supply has crap capacitors, while the mainboard is home to several Nichicon caps, just at the DC input. Low-ESR, high-ripple HD series, no less.

Speaking of DC input - surprising to see one of the wires sheared right off, from the factory. At least in the Konnekt 48, that's the "PowerGood" signal from the power supply; there's one pad in that group of five that has no wire going to it, but that looks "factory", so it can't have been removed from there, post-manufacture.

Tucked under the small daughterboard carrying the optical ins & outs is the (by now) "ubiquitous" Texas Instruments TSB41AB2 Firewire interface. Judging by the shinier solder on its pins, one ceramic cap severely cocked-over and a few other shiny and blobbed-up solder joints, i'd say this had the chip replaced at some point. No big deal, but it's a small insight into the unit's history, i suppose.

Two of the three larger buck (step-down) converters on the board, the nearest to the mains PSU, are operated by a pair of ST L5973D chips. Nice chunky 15uH inductors and a smattering of beefy 1210 and 1812 size ceramic caps. One of the two 27uH inductors on their inputs seems to be missing a chip off its top, but i doubt that affects much, if anything.

The third buck converter is run by a TI TPS64203 controller, using a Fairchild Semiconductor FDFS2P102A P-channel MOSFET and Schottky rectifier combo. Given its proximity to the Spartan FPGA and the SRC chips, odds are this may well be powering at least part of them.

A lone National Semiconductor L1117 handles the 1.8v rail for the Dice II processing core. And that about wraps up the power section analysis "detour" - time to get back to the signal stuff.

While the optical I/O is handled directly by the Dice II, some bog-standard 74HC-series logic buffers and drives the S/PDIF ins and outs. Slightly different story for the AES/EBU section, though.

Fancy stuff - a bespoke pair of high-speed differential line driver and receiver, DS26LV31 and DS26LV32A respectively.

A quartet of Asahi Kasei Microsystems AK4112B digital interface receivers is in charge of the conversion from AES/EBU to I2S. Before looking up the datasheet, i would've thought they might be bidirectional, but since that's not the case... Perhaps they're doing the I2S to AES/EBU conversion using the FPGA, then? I see no dedicated output converters, so your guess is as good as mine.

Wrapping up the digital side of things is a Texas Instruments TLC2933A phase-locked loop. Makes sense for it to be in the vicinity of the sample rate converters, what with being involved in clocking issues.

On to the analog section, as brief as it is. One small AKM AK4385 stereo DAC, along with a dedicated ST L78L05 linear regulator - gotta love the attention to detail.

Only two opamps in sight - an ST MC33079 quad, driving the line outs, and a TI NE5532A dual for the headphone output. Interestingly enough though, in addition to the latter being a reasonable enough driver in its own right, it looks to me like they put together a pair of discrete class (A)B current buffers. Diodes biasing the bases open, 33 ohm emitter resistors - textbook stuff, but i don't recall having seen this in a very long time. More details on the concept can be found right over here (figure 1 in particular).

Right then, with the "dissection" out of the way, getting down to the actual business at hand - investigating the "coma" it's in.

First round of measurements, with only the mains power supply plugged in, but front panel (and thus, "power button", which is rather a stand-by switch) disconnected: orange wire +14v, black's ground, red wire +8v, yellow wire +14v. Pretty much in line with what the Konnekt 48 supply did, and almost "surprisingly enough", quite alive, so that can't be it.

The right-most buck converter puts out 4.95v, so that's the 5v rail. The one next to it though, not so much - a mere 0.11v. That's no good. And judging by the size, that'll likely be the 3.3v rail. Sure enough, no voltage on the input or the output of the 1.8v regulator. Buck no.3 puts out 1.2v - bang on what (simpler) FPGA's need for the core voltage.

Right, diode test shows a dead short on that voltage rail. Time to remove the inductor and see which side of that the short to ground is on.

Well... the short remained on the "output" side of the inductor. I'm not exactly what you might call "happy". Ah well, bench power supply time - feeding in 3.3v and seeing what gets warm. 

6A doesn't produce all that much, as it turns out - on the current limit, the output voltage drops to 0.31v. Heating test's not too conclusive either. The TSB41AB2 kiiiinda-sorta seems to get a smidge warmer than anything else on the board, but definitely not hot. But then again, there might be some components on the bottom of the board as well. Time to find out, wouldn't you say?

Not much at all, really, apart from supply bypass caps, some ferrite beads and diode clamps on the AES/EBU i/o. On the other hand though, someone had messed with the quartz crystal and associated caps, that are connected to the Firewire chip. This is starting to get really suspicious indeed...

Yeah, i've had it, that TSB's coming off. At least that's one variable removed from the equation. 

HA!!! Whaddya know - no more short to ground on the 3.3v rail. Right then, since i'm out of brand new TSB's, i'm scavenging one out of a(nother) MOTU 828mk2. I want this thing fixed TONIGHT!..

[10-15mins later...]

Aaaaaand we've got another live one!... 
[insert humongous sigh of relief here]