Thermal Camera Teardown - The FLIR TAU 320 by Fraser

[bktemp]

Would it be possible to use boards from the working unit to see which board is faulty?
Could it be a dead sensor? It looks like all the analogue stuff is integrated into the sensor, so if the FPGA does not receive any valid data from the sensor it may go into fault mode.
What happens if you connect the sensor from the dead unit to the good one?

MT46H8M32 is a DDR SDRAM, not an SRAM.

[Fraser]

@bktemp

Already done 

The working PCB works fine with the non-working units microbolometer PCB and sensor. Obviously the calibration and NUC data is invalid, but it works. The 'main' PCB is at fault.

Thanks for the typo correction 

Fraser

[bktemp]

Good to hear that the sensor is ok.
Debugging such a dense design is probably very difficult unless it is a rather simple analogue problem, because almost all signals are hidden on the inner layers.

Could it be corrupted NUC/calibration data?
It looks like the FPGA is running, so the main block of the SPI flash must be ok. The calibration data is probably stored at the end of the flash after the FPGA code.
If you can't find any hardware fault, you could try to copy the SPI flash from the working unit.

[Fraser]

As added information for others working on these cameras.

Even without the Microbolometer pcb connected, the TAU PCB will output a composite video image with the FLIR LOGO in the top right corner. The green FFC event square also appears in the top left corner after initial power on. This camera board does not seem to care whether or not it is receiving information from the microbolometer and, unlike my PM series cameras, there is nothing on the microbolometer PCB needed for completion of boot. 

So in summary

1. Disconnect the Microbolometer PCB ribbon from the main PCB connector
2. Connect an NTSC monitor between Pin 19 of the ADV7127 video DAC and the 0V rail
3. Apply +5V to the PCB via the Hirose connector or direct to the power input filtering chain
4. The TAU PCB will boot and draw 134mA at 5.0V
5.The FLIR Splash screen will appear
6. A plain grey screen will appear with the white FLIR Logo in the top right corner

If you do not get the same as above, your main PCB is likely not operating correctly

Fraser

[Fraser]

@bktemp,

Sadly I am not certain that the FPGA is working correctly, or even running 

I did think the TAU was generating the colour bar display whilst it was in the HS324 camera, but sadly that is incorrect. The colour bar display is in fact being generated by the later video stages that are fed with video from the TAU. I believe the video stage generates a colour bar image until a valid video signal is presented to it. As the faulty TAU is not producing any video, the colour bar display remains active.

Fraser

[Fraser]

Next steps will be to recheck all power supply rails after I fit parts to replace those I removed from the 9V supply. I will then start the usual checks on clocks, reset lines, bus activity etc. The FPGA Reset line is connected to an LED so I can see that the RESET is not asserted.

Compared to other cameras that I have worked on, this is relatively simple in some ways. Far fewer components to check. That said, it is high density and access to some IC pins will be challenging, if not impossible.

Fraser

[Kilrah]

I would think the FPGA is running if you say the current drawn by both boards is identical. An unconfigured FPGA will typically have a significantly lower consumption than one that's running.

Since no video seems to be the main fault I'd start checking the video path starting from the output...


EDIT: Forgot the shutter issue... 

[Fraser]

Kilrah,

Thank you. I completely missed that an unconfigured FPGA will draw less current. Good spot chaps and very helpful.

Yes the lack of the FFC flag activation is a fly in the ointment. Otherwise I would be looking at the video DAC for a failure.

The FPGA may be configured, but in some form of HALT due to an issue that I have yet to discover. I shall have to spend a little time on this little core and attempt RS232 communications with it.

Thanks for the input guys, very much appreciated 

Fraser

[Fraser]

Some 'quick & dirty' screen captures of the good TAU PCB running with just a 5V supply and Pin 19 of the video DAC connected to a test monitor. Note that the test monitor has a single white pixel fault that is nothing to do with the TAU. The 'AV2' annotation is generated by the test monitor so I know which input channel I am monitoring.

Fraser

[Fraser]

More of my TAU diagnostic 'working notes'. Yep they are ugly, but good enough for what I need during a repair 

This time identification of the power supplies used in the TAU 320 main PCB. Other separately regulated supplies exist on the Microbolometer PCB.

I have indicated the inductors associated with each supply rail as they are easy to identify and you just test the voltage at the inductor output.
On the boost and buck converters the layout of the circuit is very simple and clear so the schottky rectifier diodes and smoothing capacitors may also be easily identified.

The indicated supply voltage are theoretical values so I have tested my working PCB with the microbolometer disconnected and just a 5V0 supply connected to it. The 'real world' voltages for each rail are as follows:

1V2 Rail = 1V210
1V8 Rail = 1V825
2V5 Rail = 2V512
3V3 Rail = 3V348
3V6 Rail = 3V633
5V0 Rail = 5V000 (as set on Power Supply)
12V Rail = 11V59

On my faulty PCB there was a 9V0 Rail in place of the 12V0 Rail

9V0 Rail = 9V30

All voltages measured with a Fluke 87 III multimeter

PCB current consumption = 134mA Stable

Fraser

[bktemp]

It looks like most pins of the BGA packages have a via next to it. So it should be easy to look at the CLK pins of the DDR SDRAM. If the differential clock is present, the FPGA is running.
You can also check the clock line and the data lines at the video dac.
Since all voltage rails are fine, my guess would still be corrupted SPI flash data, because there is not much else on the PCB that can go wrong except from broken solder joints or vias.

[Fraser]

Hi bktemp,

Many thanks again for your help on this. I would be the first to admit that working with FPGA chips is a bit of a nightmare for me. They are great for the designer and integration, but a nightmare to diagnose faults as they do so much inside that little BGA package, with no way to know the configuration unless you are the designer. What I do like about them though, is the fact that they are configured on boot rather than a custom one time programmed device or mask programmed device. If one fails, you just put another one in and the Flash chip provides the configuration at boot Well that is my limited experience with them anyway. Far better than custom programmed chips IMHO.

If the FLASH has been corrupted, I may have a challenge on my hands as that will contain the essential Microbolometer dead pixel and NUC data.

I hope to have some time to 'play' this afternoon so may make some progress with the diagnostics. It is fortunate that I have a working unit that can provide, hopefully, useful reference data to compare with the faulty PCB.

Fraser

[Fraser]

An updated and corrected TAU 320 block diagram plus the schematic of the +5V supply input filtering.

With regard to the +5V input, it has no reverse polarity protection and FLIR warn of this in the User Manual. I would have expected a diode across the 5V rail and a very fast IC protection fuse in line with the supply input. This camera is not exactly cheap !

Fraser

[Fraser]

OK, I have 'proof of life' from the faulty TAU core 

She is currently failing to initiate the two FFC events that sort out the image and vertical lines in it. There is no input occurring to the motor FFC Flag driver IC, and no green square in the top left of the image to show the FFC event is about to start.

Pictures attached. Note the Indigo Logo on this TAU at boot. You will recall that I said it was an Indigo design and not from the FLIR designers.

More when I have some spare time.

Fraser

[Fraser]

My hand in front of the lens. All seems to be working except FFC.

[Chanc3]

My hand in front of the lens. All seems to be working except FFC.

I don't mean to sound silly or anything, but is the shutter stuck?

[Fraser]

Hi Chanc3,

No, sadly not so simple. I attached the whole optical and microbolometer assembly to the good PCB and it worked fine, including the shutter. There is also the matter of the green FFC warning square not appearing and no signal to the motor driver to actuate the flag.

I have just received some USB to 3V3 UART adapters that use the CP2102 chip. This is what lives inside the FLIR USB adapter module. I will make a DIY version of the FLIR adapter module and use the FLIR GUI to see how the TAU is configured. etc. Sadly I have virtually no spare time at the moment.

Fraser

[Fraser]

For anyone wanting to connect to a TAU core without the FLIR VPC module, here is the information you need.

The FLIR VPC  provides the following facilities to the user:

1. 5V power for camera from USB source or external 5V power supply. Current at boot is ~500mA dropping to a nominal 180mA

2. USB to Pseudo 'RS232' conversion, providing the required 3V3 levels for the core.

3. Analogue Composite video output NTSC or PAL depending upon core settings.


So what 'lives' inside the VPC ?

In truth I do not know as I have not seen one up close. I do know that it contains a CP2102 USB to UART converter that provides a 3V3 TTL output. The VPC may contain some power input protection, I would hope so for what it costs.

To build your own VPC you will need the following parts

1. Hirose DF12-50DP-0.5 header connector (costs less tha £1 from Farnell)
2. Some form of PCB on to which to attach Hirose connector. Pin pitch is 0.5mm
3. CP2102 converter PCB from ebay (costs less than £2.00)
4. Video connector of your choice
5. Micro USB cable
6. Micro USB Power supplu 5V, 1A max (optional)

The pins that must be connected to the core connector:

Pin   Function

1     RS232 TX
2     RS232 RX
5     D Gnd for RS232
43   Video out High
44   Video out Low
47   0V
49   0V
48   5V
50   5V

Note that the TAU core provides a Hi-Z analogue video output that is intended to be terminated with 150 Ohms or 75 Ohms depending upon user requirements.

I would add a protection diode across the 5V supply rail and a series high speed fuse to offer some protection to revers polarity errors.

All this costs a few pounds and is pretty easy to build. The official VPC costs around £150 in the UK and has to be ordered.

The TAU GUI software for testing and configuration is free and available from the FLIR download page.

http://www.flir.com/cores/display/?id=51880

The software comes with the CP2102 driver software and installs it after the GUI installation is complete. Power can come from the USB port so no other power supply is needed. The CP2102 PCB that I bought from ebay includes a 5V output pin for just such use.

I include pictures of the FLIR VPC, CP2102 PCB, Hirose DF12 connector information and an extract from the TAU manual detailing the interface and pin layout.

UPDATE: Note that the TAU uses inverse UART polarisation. Tht is to say, Logic 0 = +3V3 and Logic 1 = 0V. You need to place a 74HC04 inveter IC in line with the RX and TX lines if using a standard USB to CP2102 adapter. Take the 3V3 supply for the inverter from the CP2102 PCB. All will work well then.

Fraser

[Fraser]

I have all the parts I need to make a DIY VPC adapter. see pictures.

Surprisingly the prototyping PCB has the exact same mounting hole spacing as the FLIR PS32 adapter PCB. I will join teh two PCB's together with some stand-offs and 2.5mm screws.

I decided to X-Ray image the PS32 adpater PCB to see how it is configured. It is a hybrid FPC/FR4 PCB so I was looking for hidden tracks within the layers. Those images will be in my next post. It sure makes life easier when you can X-Ray image a PCB.

Fraser

[Fraser]

Both sides of the FLIR PS32 adapter PCB that connects the TAU core to the controller PCB. I am 'borrowing' it from my spares donor PS32 unit to use on the faulty TAU core for tests. The adapter will be returned to the donor once I have finished with it.

Fraser

[Fraser]

X-Ray images of the PS32 adapter PCB showing all the tracks and their connectivity.

Fraser

[Fraser]

X-Ray image of the PS32 adapter FPC but I will not be connecting to the edge contacts.

Fraser

[Fraser]

I had a few minutes this evening to knock together the little TAU test jig. It is a simple little assembly so did not take long.

I tested it with the TAU PCB and the FLIR GUI for TAU cameras. The result...... communictions with the core could not be established. The CP2102 USB-RS232 bridge is recognised and working, but the camera is not responding to the GUI initialisation sequence. Tried different baud rates, tried reversing TX and RX lines, no dice. tried the working core and that behaves the same. Why can't things be simple eh 

I will have to scope the TX and RX lines to see what is going on. I will also try direct commands via Hyperterminal to see if any response can be obtained from the core. No time or inclination to do that tonight so it will have to wait.

Such is life 

Fraser

[Fraser]

Jig with TAU core fitted.

[Fraser]

I may have found the reason for the lack of communication. A chap on another forum was having the same trouble with a TAU and FLIR told him the logic on the TAU serial port is Logic 0 = 3V3  and Logic 1 = 0V.  It is inverse to normal unbuffered UART Logic. I will have to graft a 74HC04 Hex inverter in the serial comm's lines. What a PITA 

Update: I just found the interface and software design documents for the later TAU2 camera. That camera has auto detection of logic polarity but always defaults to inverse logic in the first instance. That kind of confirms that the TAU runs inverse logic. The earlier TAU does not have auto polarity detection, hence why I am having problems. Why have FLIR done this you may wonder....... well they say the I/O on the TAU follows the RS232 protocol except in terms of voltage levels. In RS232, any voltage more negative than -3V is a Logic 1, and any voltage more positive than +3V is a Logic 0. Shifting those voltage into the single polarity domain above 0V and changing to 0/3V3 logic levels results in FLIR's inverted UART logic on their camera I/O of L1=0V and L0=+3V3.

Powering a 74HC04 from the 3V3 supply rail should sort things out as that IC works down to a Vcc of +2.0V. It will just invert the TX and RX lines polarity.

Fraser