OPA549 permanent thermal shutdown

First and foremost get all of the failures back here for Failure Analysis in parallel with our applications discussion.  Go back through wherever you purchased the OPA549s from to process the Failure analysis.  Mention you have communicated with the factory and my recommendation is to return parts. 

In parallel:

1) what type of thermal material are you using between the power package and your heatsink?  Vendor and part number is preferred.

2) Confirm M4 machine screws for mounting power package to heatsink?

3) Mounting torque used on M4 machine screws?

4) Are you using an insulating shoulder washer on the bolt?  Vendor and part number preferred.

5) Thickness of aluminum heatsink is 0.750 inches?

6) The OPA549 Power Op Amp leads are soldered in the PCB first and then the heatsink is bolted to the assembled PCB?

7) A quick scan of your schematic indicates probable V-I motor control circuit.  Command voltage in yields motor current out in a bridge configuration? Can you get inductance and resistance of motor loads?  Attached is a PowerPoint that discusses stability issues with inductive loads and V-I circuits.   If you get Inductance/resistance we can help analyze for guaranteed by design stable operation.

Tim - thank you for the quick response and PPT. It will take me some time to digest that and to gather all the answers.  My group at JHU (Baltimore) did the high-level design. We used an outside consultant (with 25+ yrs experience) for the detailed design and layout. The board assembly was done by a company in Ohio. The initial testing was done at WPI (in MA), and the failed boards were sent to us for troubleshooting.

I am happy to work with you to improve our design, but I should mention that these op amps never drove a motor. Our test process (also used at WPI) is to first connect a new board to a test board that has a 10 Ohm power resistor instead of a motor (that board design is also on GitHub).  We then run a series of tests that exercise all inputs and outputs and the failures were detected during this testing.

Here are the answers that I can provide now:

1) It is heat sink compound type 120. I will have to get details from the company in Ohio.

2) Hmmm.  Will have to investigate. Heat sink drawing says M4, BOM says 4-40.

4) Since our V- is GND, we did not see the need to insulate. We expected op amp tab to be connected to V- (as specified in datasheet), but it appears to not be connected.  Should we connect our heat sink to GND on the PCB?

5) Yes

6) The op amps are first connected to the heat sink and then the leads are soldered. The company used a selective solder machine this time.  Previous runs used wave solder.  In both cases, solder is at 260 degC.

7) Yes, it is V-I motor control. We are driving small Maxon motors in a robot.  Largest motor has R=7.73 Ohms, L=1.31 mH.  In some cases, we drive two of these motors in parallel.  We use a +24V supply.  In the future, we may drive larger motors. A few years ago, we did some testing with a larger Pittman motor.

Mounting torque on bolt will be important.  I am researching with original IC designer if all of the tabs tied together and floating is a good idea.  Should be able to update you on this tomorrow.  Please get failing parts back on an FA cycle from where you purchased them as fast as possible and put my name as reference on the FA so I can help when the parts are returned to TI, Tucson Design Center.  I can do some analysis I will share with you on the circuit Loop Stability. I cannot tell by inspection if your current circuit is marginally stable.  Marginal stability on power op amp circuits is bad and you do not want it in production unless you like random destruction...  Will share the analysis with you once it is complete.  

I got more information from the assembly company:

The torque on the screws is 12 in-lbs.

The thermal compound is Wakefield Thermal Solutions PN: 120-5, SILICON GREASE 5 OZ TUBE.
They apply a small bead to the parts, then use a small metal squeegee to leave a thin film
of grease on the part. It is an eyeball thing with the tech doing the job where they need
to ensure that a thin, uniform coating covers the entire tab.

They are using M4 screws to mount the op amps to the heat sink (I need to update the BOM).

I will convey information about FA cycle to them.  Thank you for your help!

You may want to consider Loctite Thermstrate which is really nice to work with and not messy like grease.  It is a "This product is supplied as a dry compound coated onto an aluminum substrate. The compound flows at the phase change temperature and conforms to the surface features of the heat sink and component. Upon flow, and in conjunction with component mounting pressure air is expelled from the interface, reducing thermal impedance and enabling the material to perform as a highly efficient thermal transfer material. Loctite Thermstrate is supplied as die-cut preforms to match a wide variety of electronic components. Custom parts are also available upon request."

Also get me a drawing of the mounting bolt with detailed specs on is head diameter, etc.  I want to pursue mounting procedures in parallel.

Sooner the better on those FA parts....

Thank you for the reference to Loctite Thermstrate.

We mount the OPA549 to the heat sink using an M4 screw.  It is McMaster-Carr PN: 92005A218, Metric Pan Head Phillips Machine Screw, Zinc-Plated Steel, M4 Size, 8mm Length, .7mm Pitch.  The head diameter is 8 mm,  and the head thickness is 3.1 mm.  The drawing is available here:  http://www.mcmaster.com/#92005a218/=s6gfs5

The assembly company has contacted Digikey to start the FA process.  I don't know the current status.

I took the liberty of building up your circuit in our free SPICE simulator TINA-TI which can be downloaded here:

http://www.ti.com/tool/tina-ti

I attach the schematic here for one coil and checked loop gain stability which looks good.  The next post is the TINA-TI schematic for loop gain stability check for Two Coils in parallel and it looks good as well.

Will continue to research mechanical mounting analysis while we wait to get parts back on an FA.

Post with two coils in parallel for loop gain stability.

1) I checked with IC Designer and your arrangement of a "floating heatsink" is okay since all parts are running from same V- supply of ground.  If there is ever another voltage that touches this heatsink it will take out the power op amps.  If the heatsink were grounded on each end then if another voltage touched the heatsink it would pull that supply to ground but the power op amps would be fine. If you have no concerns of another voltage ever touching the floating heatsink then you are okay.

2) We have serious concerns based on your 4mm mounting bolt. See attached PowerPoint and let me know if any questions.

Wow.  An interesting theory that the 4 mm mounting bolt will cause an interference fit that "can cause undo [undue] mechanical stresses and concerns of power tab deformation during mounting which could cause die de-lamination."  Have you seen this before?  Given that we haven't found any other likely culprit for the failures, this could be it.  Would it be possible to determine whether this was the actual cause in the Failure Analysis?

Regarding the "floating heatsink", it was not our intention to float it.  The OPA549 datasheet claimed that it was connected to V-.  It isn't too likely for the heatsink to come into contact with another voltage, but I would prefer to guard against that.  For example, I assume that accidentally shorting the heatsink to V+ would not be good.

FA analysis will help us determine the root cause of the failure.  It is critical that you eliminate this now known mechanical variable of an interference fit in the mounting hole of the tab caused by 4mm bolts. Yes, I have seen mechanical stresses on heatsink mounting of the power tab cause die damage. If the power die becomes de-laminated it will not conduct heat to the power tab and thermal shutdown will occur quickly.  

Yes, accidently shorting the heatsink to V+ could take out all of the op amps on the heatsink.  My preference would be to ground the heatsink on each end.  If not practical at least on one end.

The distributor (Digikey) has finally sent the parts to TI for Failure Analysis.  This process took much longer than I expected!