Managing user disappointment

[JÞB]

Good day all.  I've got a tricky problem with a test that will never be user friendly and have no means to change the design of the UUT.  I will need to manage the user's. 

Functional test system for a circuit board where DFx was not high priority.  The test is simple.

Establish a condition,

Loop

       command uP to get an ADC value

       Ask for that ADC Value

       Display to user

Iterate

Loop rate is 750mS and the potentiometer the user is manually adjusting is very fine.  Worse, the uP will reboot if the temp sensor on the board gets too hot and the ajdustment affects this temperature.  Meaning of course the communications with the uP go bye-bye the test cannot pass and the UUT will need to cool before a retest can be even attempted.

I cannot think of a way to make this a "push the button, get a banana" rock simple test step.

What have you done in this type of situation?

[Ray.R]

Is there any way to automate the potentiometer?

By how much does a small change of the potentiometer affect the values which are displayed?  Could you show the mean of the last 5 measurements?  AT 750ms, the operator won't be able to tack the changing values, but slowing it down to 2 or 3 secs might be ok.. Would an avaerage over that period work?

How skilled is the operator?

Can you provide an actual scenario?

[crossrulz]

Is this an actual test or trying to find a select value?  If trying to find a select, can you figure it out at a lower subassembly level without the uP?

These type of things always annoyed me.  And if your customer is like my former employer, the test engineers are not included in the design process until it is way to late.

[herrlin]

I have assumed that the user is making adjustments in parallel to your loop.

Could you first determine how much a certain amount of adjustment effects the value then better advise the user how much adjustment is needed with the goal of avoiding an over heat condition (conservative adjustment calculations).

Example

Get ADC Value

Instruct user to adjust potentiometer X

Get ADC Value

For (until value is acceptable)

   Determine ration of adjustment to value change

   Calculate needed amount of adjustment

   Tell user to make calculated adjustment

End

[jcarmody]

---

@Ray.R wrote:

Is there any way to automate the potentiometer? [...]

---

---

@crossrulz wrote:

Is this an actual test or trying to find a select value? [...]

---

 Digital potentiometer?  (assuming that the potentiometer isn't part of the UUT)

[Ray.R]

---

@jcarmody wrote:

---

@Ray.R wrote:

Is there any way to automate the potentiometer? [...]

---

---

@crossrulz wrote:

Is this an actual test or trying to find a select value? [...]

---

 Digital potentiometer?  (assuming that the potentiometer isn't part of the UUT)

---

That's what I had in mind.

[JÞB]

Thanks for your points.

Yes, a digital pot would be nice.  There really is a closable loop that would be able to set it.  Design change will not be happening.

The "Good" range the seams to be about 10 degrees on a one turn pot so faster updates would actually be ideal (The user is looking at data that is up to nearly a second old while adjusting)  I think I'll "Flash" the entire panel at update time to give the user a sort of cadence for when he gets an updated value. and handle any overtemp shutdowns with a brute force "The uP has rebooted due to an overtemp.  Turn the pot CCW, press cancel and try again when the UUT cools.  Sorry for the inconvieniance!:)"

Eventually some briliant line operator will suggest adding a "turn pot fully CCW before placing part in fixture" step and a "turn slowly" warning on the fixture.  I hope they forgive me!

[johnsold]

Jeff,

Can you put some resistance in series with the pot wiper and connect that to a clamp circuit (under software control) which whould limit how much the  output could change?  This could create other problems for the user (as in "I turned the pot all the way and nothing happened!"), but might prevent the shutdowns.  Then you could warn the operator to try again slowly.

Lynn

[JÞB]

Lynn,  Nice idea but Unfortunately not,  I have to test the board the way it is and yes, operator error can make the board untestable.  There will need to be operator training- its an unfortunate design decision  The customer will accept the solution presented and admits culpability but the operators are not going to like it.  I'll just put some other name in the vi info.  You may hear rumors about that idiot "John Gault" who designed the test.

[crossrulz]

---

@JÞB wrote:

You may hear rumors about that idiot "John Gault" who designed the test.

---

Or you could put "Pointy-Haired Boss"