FPGA oscillator

[indyld]

Not sure where to put this, a bit off piste for Lighting even though it's a lighting console.

Currently testing a mobo that doesn't POST and decided to probe the xtal next to the FPGA chip. Xtal says 24.5 on it. 

I get a sine wave of about 1.2v p2p on one leg. 3v3 solid DC on two others plus GND on the last one. 

Hard to see if any caps related on the back of the board. Also, my crappy handheld Chinese scope that I usually just use to prove signal says 4.5 MHz.

Does all this mean the FPGA is running ok, you reckon?

I've got other avenues to look down too but clock activity seems a reasonable start point. 

Edited November 30, 2022 by indyld

[kgallen]

What's the Xilinx part type - e.g. Spartan or Virtex, will be a number like X3Sxxxxx or X2Vxxx for example.

Without more info probably what you're observing is ok assuming your scope is quite low bandwidth and loading the crystal oscillator quite a bit!

[indyld]

Spartan XC3S50A FTG256AGQ 1345

[kgallen]

So assuming the xtal does go to the FPGA then that Spartan XC3S50A supports I/O at 1.2V, 1.5V, 1.8V and 3.3V. As it’s a programmable device we can’t know which I/O type is selected without the source code. But since you measured 1.2Vpp from the xtal that will be above and below the logic thresholds for most of those I/O options. Gut feel is it’s ok. 

https://docs.xilinx.com/v/u/en-US/ds529

Edited November 30, 2022 by kgallen

[indyld]

Thanks. Am wondering if the array is relevant to the boot issue or just deals with the considerable external IO on the board. I think this one may be working at 1.2v based on other observations.

If, say, the array isn't being setup by the nearby PROM for whatever reason then I dunno if that might prevent any boot activity at all e.g the power up button running through that IO. (The button seemingly does nothing other than shorting its 3v3 to ground quite happily: no beeps, no change in fan, no change in current draw etc.)

It's all very mysterious and I'm working on the basis that I'm not gonna find out an easier way than poking around. My experience of the maker has not been of the 'speak to Dave, he designed the thing' variety. Still, there's time to eat my words yet. 

Edited November 30, 2022 by indyld

[kgallen]

Quite likely the FPGA ‘programme good’ output could go to any micro so a failed FPGA config could be related. Usually they use a small Xilinx serial PROM for FPGA config and there are extra signals you could probe (take a look at that data sheet linked above there will be a config diagram). Without schematics though any debug will be quite tricky. 

[indyld]

Agreed. I'm used to working without schematics but given that I don't have access to the programming either if the issue is caused by dodgy config then I think I'm SOL on this one. First external observation on the desk was a mangled USB but hopes for this being a straight forward power management fix went when all standby supplies appeared fine on the board. Not really been able to trace the power button terribly far, probably disappears down a via under another pad. At that point, I think we may be in the realm of interested poking rather than expecting a solution.

The board has plenty of status LEDS on it, mostly all lit positively. There are five marked near the FPGA: 1.2v, 1 , 2, 3, and one marked Done. Am trying to get info from the maker what the current LED state tells me, if anything. 1.2v is lit, 1 & 2 are flashing, 3 is lit, Done is not.  Might be significant, might not be. On to slightly less esoteric things tomorrow I think. 

Will trawl through the datasheet and try to make at least head if not tail of some of it. 🙂

Edited November 30, 2022 by indyld

[kgallen]

Done led sounds promising for fpga config. It will be a dedicated output pin on the fpga (or serial prom I can’t remember which).

Edited December 1, 2022 by kgallen

[indyld]

Had a quick poke around the PROM before starting other jobs this morning. There are a couple of likely looking candidates around the FPGA, but I think the relevant one is a ST M25P10 (on the right in the image). Can't seem to see anything at all on the CLK of that IC and the Q output seems like it sends out 90KHz of 11mVpp garbage, the drops dead for while, then starts again. 

The other nearby PROM is marked as an ATMLH150 which I failed to find a datasheet on but doesn't appear as connected to the FGPA. It's also near the SATA port components. The ST also appears to be hooked in with all the stuff around the programming header.

Not sure what use I can make of this information unless the manufacturer blows cover really, but it's interesting to think about nonetheless. 

Edited December 1, 2022 by indyld

[timsabre]

Rewinding a bit, when you say it doesn't POST, how are you determining that? Could it be a display output problem or similar?

[indyld]

After removing it from the console, I've not even hooked it up to a display as it doesn't even try to boot. The board is supplied by single 12v from the PSU rather than a more traditional multi voltage and the current on the 12v remains unchanged, along with no beeps, CPU fan etc. stays low speed or other indications of starting POST. Symptoms are the same, either on or removed from the rest of the assembly and also with or without anything external, RAM, GPU.  When on the console, power applied to the PSU, all the on and off board fans run (which I am not sure is how it's meant to be to be honest) and the CPU fan goes round at a steady tickover speed. The many status LEDs mostly light up, certainly I've not found a power rail that isn't doing what it's meant to be.

The board on its own draws about 900mA with the GPU and CPU boards in place, and about 500mA with only the CPU. There's nothing particularly hot but nearly 1A seems like a lot of standby to me! It may be just how it is.

Nothing changes when the on-board power button pulls low.

I guess that next thing to wonder would be where the SPI clock is made. There's an MCU on the CPU board that I haven't look much into, 3 xtals on the main board: near the FGPA, near the networking controllers, and near a USB controller.

Edited December 1, 2022 by indyld

[indyld]

Visual aid for those interested. The gaps are the missing GPU and SSD. Their presence or otherwise made no difference. The RAM sits on the main chipset daughterboard under the fan and is also currently removed.

All off board stuff is various headers to 'things' or USB. 

[indyld]

Heard back from tech support, the FPGA LEDs status matches a working board so at least that wonder is dealt with. I think one issue is that they don't actually know very much more about the board. It's common these days that even the best of company tech support have to say 'Sorry, we aren't given any component level information on that'. 

[richardash1981]

If this is a more or less PC-like design (and from the pictures and what has been said it seems to be), I would not be surprised to find that the x86 CPU is the first boot device, and things like the FPGA get initialised much later, essentially as peripherals?

From that starting point, I wouldn't expect to get far without the GPU/SSD/RAM and would be looking for BIOS/UEFI PROM, and associated SRAM with a coil cell for memory backup ...

[indyld]

It is basically a PC, it's just all very original and a lot of the design isn't as familiar as many high production PC / laptop boards. The thing that I've not found definitely is what actually controls the boot which is usually either SIO, SMC, PCH,  Southbridge or whatever. This is why I wondered if the FPGA was what did that in this case. Not just because of the position on the board near the power button input and what is probably the BIOS PROM. It could also be the MCU on the CPU board but something is lighting the large quantity of status LEDs and one or other seems likely. 

I've tried all the usual things to reset the BIOS/UEFI. A mobo will at least try to start with very little on it and removing potential faulty extraneous things like RAM is a common diagnostic step. 

Tracing the boot switch line has proved particularly difficult even by today's multilayer, stick-it-wherever standards but if it disappears into a 256 pad BGA, it's gonna hard to find the fault without schematic or even just a boardview. 

The question at the moment is mostly what is doing the job of the SIO.

This isn't solved by asking the manufacturer, as their tech support don't have access to such detail. This situation is pretty normal these days, I recently spoke to a UPS maker as I had a query about the main board. The standard answer is "dunno, we just change the board or, on the lower end models just write the device off". 

Someone somewhere knows how these things work, but good luck trying to pin them down. It's why I do so much business fixing all these mystery devices. In this case, for the owners its either I fix it or it doesn't get fixed and they go without. I'm often the last stop on the route to the trash. 

 

 

 

 

Edited December 2, 2022 by indyld