XMOS boot question

[mark03]

I recently bought a non-working Behringer UMC404HD (fancy USB audio interface) and have been trying to figure out what's wrong with it.  It does not enumerate when connected to USB, so I opened it up and discovered the "brains" are an XMOS 8U6C5.  It's supposed to boot from a program image in SPI flash, but that's not happening.

Poking around at the supply voltages and comparing with the reference design in the datasheet, the 5V supply is fine, and 3.3V is provided by an LDO to the VDDIO balls, also fine.  There are two on-board synchronous bucks for the core voltages.  Neither of these are running.  So the chip has 5 and 3.3 volts but that's it, no VCORE.

Probing on the debug header, RST_N is low (should have internal pull-up).  The reference design shows MSEL on the header connected to the MODE2 and MODE3 pins; I can't verify this on my board (BGA package), but if true, MODE2/3 are also low, which implies "boot from JTAG."  These pins should also have internal pull-ups.

My question, for anyone familiar with XMOS, is where in the power-on / boot sequence do the DC-DC switchers normally start?  Is the chip able to start loading the firmware without them, and then expects to see the right bits set to enable the switchers?  Or should it happen as soon as 5V is applied?

Or how about the oscillator?  If the 24-MHz crystal has failed, would the switchers still start up, or no?

My hunch is this is a lost cause---chip is either fried or suffered a BGA assembly defect.  But I might spend one more hour poking around if there are some easy things to check.

[maginnovision]

The 3.3v should start and when ~90% the 1v0 should start up. This is done at the hardware level, no matter what 1v0 should be coming up. After 1v0 is up reset should go high and the thing should boot. So find out why you have no 1v0. If the chip is fried it's possible 1v0 isn't coming up because it's dead. If you scope it you should see something happen at the output cap either way. 5v0 won't feed the XMOS chip, only the io and core switchers. At least in context of the actual MCU.

EDIT: What I do on my boards is have 3.3V power directly from 5V ext or an on board reg. that feeds a divider to enable the 1v0 and that is fed to a 6904 reset supervisor. The 6904 controls the reset line. Clock should be running the entire time. Also, 1V0 is core voltage, and other rails fed to it are just for IO and they should all be good when 1V0 comes good.

Sorry, read this originally on my phone and missed the P/N. I have one on hand I'll check.


Within 3ms of 3v3 going high Clock is running. 1V0 goes good after ~25ms. 1V8 goes good after ~1.2ms. At ~32ms I first see activity at the ROM. Shorting the crystal both buck converters are running, 1V8 and 1V0.
So it would seem 3V3-5V0 -> 1V8 -> Clock -> 1V0 -> ROM. The one thing I did find is RST_N goes high ~30 microseconds after 3V3 is good. That's the core booting from OTP. VSUP on this board is 5V0 and VDD_IO is 3V3.

[mark03]

Thanks.  So it sounds like the bucks are both driven from an internal RC oscillator.  I'm surprised the "core" can boot before the core voltage is available.  Perhaps there is an internal linear regulator to bridge the part until the DC-DC supplies are stable?

Anyway, it sounds like there is not much else to check.  If the DC-DC stuff is dead, the part is dead, apparently.  Replacements are ~ $16 at digikey, it's a BGA, and I'm already $50 in the hole, so I think I'll chalk this one up to poor judgment 

[maginnovision]

Yea, it's probably not worth it. I think the idea with the xs1 parts was to offer a lot on chip. So it takes common 5v0 or 3v3 at input and generates core voltages with couple of passives, on chip reset supervisor, has an ADC, internal crystal drive... All things missing from xs2 parts(and xvf as far as I know). So now the chips require more attention to detail but when most stuff that will fail is external you can actually repair them. Oh well, hopefully you have better luck with future purchases.

[Ferrograph]

Hi Did you even get to the bottom of this?

I have exact same UMC404HD but the XMOS is cooked (visibly too) - possibly due to the gottcha power connector on these being reverse polarity.

It looks like there is external flash on the PCB so Im hoping it will boot from that. Got a chip on its way.

[mark03]

No, I decided to scrap the thing for parts.  Spending any more time just wasn't worth it.  I see these boxes are now going for $200+, probably going to folks with fancy home videoconferencing setups.  I'll wait and pick one up when they drop below $100 again.

[MarviN_SSA]

Hi Did you even get to the bottom of this?

I have exact same UMC404HD but the XMOS is cooked (visibly too) - possibly due to the gottcha power connector on these being reverse polarity.

It looks like there is external flash on the PCB so Im hoping it will boot from that. Got a chip on its way.

Hi Mate, did you got it fixed? I`m facing the same here in brazil and i can`t find a place to buy this chip, can you help me with that? Thanks in advance!