Faulty medical X-Ray imager

[TerraHertz]

So here's a Hamamatsu C10900D-05 X-Ray imager. It's faulty, and is out of a very expensive machine that the owners would like to get working again. But there is no service support available from the supplier, and apparently no schematics or service info available from Hamamatsu. (I'm told, have yet to verify for myself.)
There is a specs sheet (attached.)

So as it stands it's a dead loss. I offered to have a look at it. Can't get any worse.

The unit consists of two layers, screwed together. One is the drive electronics, on a conventional PCB. From that a single connector joins to the sensor structure. That has a thick PCB, and a protective carbon-fiber cover glued to it. There's a solid metal plate between the two boards, presumably to shield the electronics from the X-Rays.
From first impression it seems I won't get to see the actual sensor plane without destructively opening it.

I haven't disassembled yet, beyond taking the metal protective cover off the drive PCB.
Right next to the interconnect, there's a row of 8 AD9220 AtoD converters.

The unit fails self test mode in the machine, giving the image cal_fail.jpg
This *should* be a solid dark gray square field at top, and a light gray field at bottom. I gather they are results of reading the square sensor array at different biases (no X-Rays.)

The white horizontal stripe in the top square is the fault. Shouldn't be there.
I notice that it seems to be one eighth of the field. Corresponding to one of the 8 AtoDs maybe?

The plan of attack is to have a better look at whatever is visible without opening the sensor plane. Then cart some test gear to the location with the machine (it's not easily movable), run test mode and see if there's any obviously wrong signals around any of the AtoDs.

Has anyone any suggestions?
What are the actual sensor devices? Charge accumulating diodes maybe? I suppose there will be some kind of CCD charge shift registers (8 of them), to clock pixel data out serially. And if something in the sensor charge shift registers is dead, then it won't be repairable.
If the fault is on the drive PCB, and can be identified (without schematics) then it probably can be fixed.

[aqibi2000]

You could move all 8 A/D one position if just stabbing in the dark

[mikeselectricstuff]

Definitely check all the voltages on the ADCs to see if there are any differences.
My money would be on a dodgy tantalum cap
 

[Manul]

While these ADCs could be a culprit, I would also look how the sensitive array (photodiode array?) is driven. That could well be a driver problem.

[mikeselectricstuff]

While these ADCs could be a culprit, I would also look how the sensitive array (photodiode array?) is driven. That could well be a driver problem.

Yes, but as there are 8 ADCs, it seems likely that it is outputting 8 parallel output streams, and the drive for each stream is likely to be common.

If you can get the panel into a continuous-reading state, it shouldn't be hard to compare the signals going into each ADC

Another suspect would be whatever it uses to connect the CCD array to the main PCB.

 

[madao]

Yes,   flat panel works with CCD . It is just a  big CCD sensor with  scintillator onboard.

 I have repaired successful few xray flat panel from Varian. Failure of CCD is not rare -> not fixable :-(

[TerraHertz]

Some more pics, after separating the driver and sensor panels.

8206 The metal plate between the two modules. Gratuitous over-CAD, making that mill show off. Why didn't they revert to straight cuts?

8208 Back side of the driver board. Sigh... more stuff that might be faulty.

8210 The accessible side of the sensor module.
sensor_03.jpg  Same, but in photoshop for tracing.
Interestingly, all the trace vias are blind, and the the only PCB holes are a grid of small non-pad holes that don't appear to be electrical.
There's a repeating pattern, times 8.  The pattern on the left in the last pic is arranged a little differently due to a layout problem. Looking at the repeated patterns of vias in a row, you'd think all the elements would be bused along some buried traces, and would come out on J1 pins.  But no! Some pins are NOT bussed, and don't go to J1 individually either. They just vanish. Maybe they are  pins of the sensor CCD(s) brought out to caps?

Not many pins of J1 are involved with the visible components. Either there are a lot of unused/power pins, or a lot of signals going to the CCD sensors via hidden traces.  Far from finished with this.

Bah. I need to go through my multimeters to find one with an audible continuity test, that's low freq enough for me to actually hear. (My hearing HF is gone. Age.)

[CaptDon]

That milling appears to be used as collumation to keep
the electron path straight to keep the X-rays 'focused'
so to speak. That is a weird milling pattern. Very sad
that no replacement parts are readily available and
would cost an insane price anyway. Got any leads to
local BioMed techs from a hospital/imaging center?
Chances are they have only done module swapping
anyway. I worked as a BioMed Tech for 5 years. The
hospital decided there would be no more component
level repairs, only module swapping. My lab now has
most of the discrete components that were destined
for the hospital dumpster. I gave away hundreds of
pounds of stuff to Vo-Tech schools and Trade Schools.
Rather see kids play with the stuff than just go directly
to the land fill.
 

[mikeselectricstuff]

If you can't fix it, it would be worth putting the sensor part number as a saved search on Ebay - you never know when stuff like this might turn up eventually.
 

[TerraHertz]

That milling appears to be used as collumation to keep
the electron path straight to keep the X-rays 'focused'
so to speak. That is a weird milling pattern.

No, the sensor is directly in the X-Ray path, and the metal plate is _behind_ the sensor, in front of
the drive electronics board. The weird milling pattern is 'just because they could.' It's actually quite
smooth, the light is reflecting off the very tiny milling imperfections.  It's not any kind of Fresnel lens
for X-Rays.

Very sad
that no replacement parts are readily available and
would cost an insane price anyway.

Very typical of corporate thinking. 'Repairable' = less profit.

Got any leads to
local BioMed techs from a hospital/imaging center?

Oh I wish. I'm sure the local hospitals dumpster all their surplus equipment,
and would recoil in accounting horror at the idea of making it available to
anyone.

Chances are they have only done module swapping
anyway. I worked as a BioMed Tech for 5 years. The
hospital decided there would be no more component
level repairs, only module swapping. My lab now has
most of the discrete components that were destined
for the hospital dumpster. I gave away hundreds of
pounds of stuff to Vo-Tech schools and Trade Schools.
Rather see kids play with the stuff than just go directly
to the land fill.

You were a rare exception.
I currently know of an unbelievable example of criminal wastage about to happen.
Large room full of a huge component stock from a major electronics repair facility.
It's probably all going to be bulldozed. But I will try to see if it can be saved. Don't
have much hope of success.

[mikeselectricstuff]

Have you asked Hamamatsu if a replacement is available, and the cost ?
 
 

[madao]

if you ask cost, pleas  sit down, before you got  answers.

I can tell: over  40k€ priece

[mikeselectricstuff]

if you ask cost, pleas  sit down, before you got  answers.

I can tell: over  40k€ priece

That doesn't surprise me, but if it's in a $400k machine....

And at that price, selling to manufacturers of expensive equipmet, they should offer a  repair service

[MadTux]

Perhaps partially erased/corrupted FPGA config EEPROM/Flash?
1000s of xray exposures probably isn't healthy for them as well.

[mikeselectricstuff]

Perhaps partially erased/corrupted FPGA config EEPROM/Flash?
1000s of xray exposures probably isn't healthy for them as well.

Pretty sure these have CRCs, so it will do nothing if the FPGA doesn't get configured

[MadTux]

Pretty sure these have CRCs, so it will do nothing if the FPGA doesn't get configured

Makes sense